Reorganize code_gen and command tests by concern

[alexkroman]

Split test files to improve organization and maintainability by grouping tests by their primary concern rather than by command.

Summary

This refactoring reorganizes the test suite to separate concerns more clearly:

• Code generation tests are now split between test_code_gen.py (serializers, snippets, transcribe rendering) and test_code_gen_stream_agent.py (stream/agent scaffolds)
• Command behavior tests are split between test_transcribe.py (output rendering, LLM transforms, sources) and test_transcribe_show_code.py (transcribe --show-code print-only behavior)
• Stream command tests are split between test_stream_command.py (source/streaming behavior) and test_stream_show_code.py (stream --show-code print-only behavior)

Key Changes

New test files:

• tests/test_code_gen_stream_agent.py — 131 lines of stream/agent scaffold code generation tests (moved from test_code_gen.py)
• tests/test_transcribe_show_code.py — 113 lines of transcribe --show-code print-only behavior tests (moved from test_transcribe.py)
• tests/test_stream_show_code.py — 109 lines of stream --show-code print-only behavior tests (moved from test_stream_command.py)

Modified source files:

• aai_cli/transcribe_exec.py — Extracted result delivery logic into new deliver_result() function and TransformOptions NamedTuple, consolidating the post-transcription routing (file output, field selection, LLM transforms, or default render) that was previously inline in the command handler
• aai_cli/commands/transcribe.py — Simplified by delegating result delivery to transcribe_exec.deliver_result() and validation to transcribe_exec.validate_out_with_llm()

Updated docstrings:

• tests/test_code_gen.py — Clarified scope (serializers, snippets, transcribe rendering only)
• tests/test_transcribe.py — Clarified scope (output rendering, LLM transforms, sources only)
• tests/test_stream_command.py — Clarified scope (source/streaming behavior only)
• aai_cli/transcribe_exec.py — Updated module docstring to reflect new focus on execution and result delivery

Implementation Details

The refactoring maintains 100% test coverage and passes all mutation gates by:

• Moving tests without modification (no behavioral changes)
• Extracting deliver_result() to consolidate post-transcription logic that was previously scattered across the command handler
• Introducing TransformOptions NamedTuple to cleanly pass LLM configuration through the delivery pipeline
• Preserving all assertions and validation logic in their new locations

https://claude.ai/code/session_014u8p2QCByGfApna8x969yV