[test](regression) skip Arrow-incompatible complex outputs

[shuke987]

What

• Skip map cast cases that intentionally produce null map keys when running through Arrow Flight SQL.
• Use Arrow-specific bitmap assertions that compare bitmap_to_string(user_id) instead of raw bitmap bytes.

Why

• Arrow Map cannot represent null keys, while the JDBC text path can render {null:...} for these cast semantics checks.
• Raw BITMAP output is protocol-specific: JDBC renders \N, while Arrow exposes binary bytes. The stable semantic assertion is bitmap_to_string.

Testing

Not run. This PR is intentionally independent of #65084; run buildall after #65084 is merged.

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[shuke987]

run buildall

[shuke987]

/review

[github-actions]

Automated review completed for PR 65182.

I did not find an evidence-backed issue that needs an inline review comment. The PR is test-only and keeps the changes focused on Arrow Flight SQL incompatibilities:

• test_cast.groovy skips only the two map-cast checks that intentionally produce null map keys when context.useArrowFlightSql() is true.
• test_aggregate_all_functions.groovy keeps the JDBC bitmap checks unchanged and adds Arrow-specific qt labels that compare bitmap_to_string(user_id) instead of the raw BITMAP column.
• The added .out blocks match the projected Arrow query shape.

Critical checkpoint conclusions:

• Goal and coverage: the patch addresses the stated Arrow Flight SQL test incompatibilities; it adds/updates regression expectations for the affected aggregate checks and skips the Arrow-incompatible map-key cases.
• Scope and focus: the modification is small and limited to regression tests and expected output.
• Concurrency/lifecycle/persistence/config/protocol: not applicable; no production FE/BE/cloud code or persisted format changes.
• Parallel paths: JDBC and Arrow paths are separated by context.useArrowFlightSql(), with existing JDBC checks retained.
• Expected output correctness: the regression framework allows skipped qt blocks by searching forward to the executed tag, so preserving the JDBC-only map_to_map2 and map_to_map3 output blocks is compatible with skipping them in Arrow mode.
• User focus: no additional user-provided focus was supplied.

Subagent conclusions:

• optimizer-rewrite: no candidates; final convergence returned NO_NEW_VALUABLE_FINDINGS.
• tests-session-config: no candidates; final convergence returned NO_NEW_VALUABLE_FINDINGS.
• No candidate became an inline comment, and no duplicate suppression was needed.

Validation:

• Reviewed the GitHub PR file list and patch for all three changed files.
• Refreshed the PR head before submission; head remained eea7146b5e6a6ac91b6c302ff865c1616cecbafe.
• Checked the PR patch for added trailing whitespace and reverse-apply consistency.
• Did not run regression tests because this checkout lacks .worktree_initialized and thirdparty/installed; validation was static.