S0070-transform-tf-8: nothing a message carries is dropped in silence

[NSchatz]

Spec: work/specs/S0070-transform-tf-8/spec.md in the umbrella repo.

What this changes

An issues list that said nothing about an AL1, a DG1 or an IN1 meant one of two very different things: the bundle carried it, or nobody looked. Two new stable issue codes tell them apart, raised once per segment occurrence that contributed nothing to a resource the returned bundle contains:

Code	Meaning
TRANSFORM_SEGMENT_NOT_EMITTED	The IG publishes a segment map for this name and nothing in the bundle carries it: a gap in this library.
TRANSFORM_SEGMENT_NO_IG_MAP	The IG publishes no segment map for this name, or the name could not be classified at all.

No segment is newly transformed. A flagged segment tells a consumer what is missing, not what it said.

The four rules that carry the safety of it

• Contribution, never inspection. An occurrence is marked as reached only where the resource it fed has just cleared the emit gate and joined the bundle. So the RXE this library counts and refuses, the orphan OBX, and a resource the gate withheld are all reported: those are the omissions a consumer most needs told.
• The name part is guarded by a shape test this library applies itself (/^[A-Z][A-Za-z0-9]{2}$/ against the value the parser published), not by excluding the parser's own marker constant. That constant is not on @cosyte/hl7's public surface, the peer range here admits any version of it, and an exclusion test fails OPEN the moment the literal moves. The shape test fails closed and selects the same set today. A test greps the shipped source to keep the literal out.
• A position carrying neither a name nor any field content raises nothing, wherever it sits, and still counts toward later ordinals so [#n] names the line the sender wrote. A line that kept its field separator carries content and IS reported.
• One issue per occurrence, numbered among all occurrences of its name. Two NK1 of which only the second is reported gives NK1[2].

Locations

NAME[k], 1-based among that name's occurrences, or [#n], the segment's position in the message, with no part of an unrecognized name reproduced. Both shapes are asserted across the whole corpus. The core-concepts doc states that these are 1-based while the per-occurrence locations the library already emitted are 0-based and have not moved, and that a damaged identifier (AL11) reads as a standard gap rather than an IG claim.

Evidence that nothing else moved

The Bundle and the full pre-existing issues list for all 23 recorded inputs were captured from this tree before a line of the diagnostic existed (commit 13750f18, first commit on this branch) and are asserted against those baselines: the Bundle byte for byte, and every pre-existing issue first, in its recorded order, with the new ones appended. That measures "no resource changed" and "no issue was reordered" against the old tree rather than against the new code.

phi-scan-overrides.md and package.json are byte-identical to their pinned content. The change is additive throughout: 0 deletions.

Gates

pnpm test 578 passed (32 files, was 524/31) - pnpm lint - pnpm typecheck - pnpm format:check - pnpm test:coverage (all directory gates green; the new module is 100/100/100/100) - pnpm phi-scan - pnpm check:no-emdash - pnpm check:no-internal-refs - pnpm check:agent-notes.