Fix: topic_segment=False no longer skips memory pipeline + defensive fixes

[keketcl]

Summary

Fix topic_segment=False not creating any memories, plus additional improvements.

Note: This PR builds on top of PR #59 (Fix: topic_segment=False), adding docstring updates, comment cleanup, and a defensive bug fix that was exposed by PR #59's original change.

Changes

From PR #59 (preserved)

Bug 1 (critical): topic_segment=False skipped entire memory pipeline — fix replaces early return with single-segment fallback.

Bug 2 (guard): force_segment=True + topic_segment=False crashed due to uninitialized segmenter — fix adds guard condition.

Added in this PR

Bug 3 (defensive init): When metadata_generate=False or text_summary=False, the extracted_results variable was never assigned, causing NameError. This was previously hidden by Bug 1's early return and exposed after PR #59's fix removed that early return.

Fix: Initialize extracted_results = [] before the conditional block.

Doc improvements

• Update add_memory docstring to reflect new behavior (single-segment fallback, no early return)
• Update Notes section: remove outdated "returns early" description
• Update Returns section: remove outdated segmentation result dict format
• Simplify inline comments to concise project-style English

Testing

Tested with topic_segment=False + metadata_generate=False + text_summary=False:

• Full pipeline executed without crash
• Returns empty result gracefully
• Module import and initialization verified in conda environment