perf(console): build the colour palette without a fork per colour

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown

[Chemaclass]

🤔 Background

Related #1285

src/console/colors.sh built its sixteen palette entries with $(bashunit::sgr N) — a subshell fork each, at source time, on every invocation including --version and --help. It was the single largest cost in a cold start, and it outlived the #801–#851 fork campaign because a PATH-shim census cannot see a subshell that execs no binary.

💡 Changes

• The palette now comes from the _BASHUNIT_SGR_OUT return slot that bash-style.md already prescribes; the public echoing bashunit::sgr stays as a thin wrapper, so its contract is unchanged
• Cold start drops ~9ms, about 15% — interleaved A/B, 3 rounds of 200 invocations in one tree, every round faster
• New cold-start guard counts the function name in the trace, since the binary-path patterns the neighbouring guards use are blind to this fork class
• perf-fork-budget.md records the blind spot, and replaces its "sourcing src/ is irreducible" claim with a measured per-source-file breakdown