fix(links): use pydash.get + decode body to fix analyze chain (CON-573)

[pavanputhra]

Summary

• Closes CON-573 — TypeError: string indices must be integers, not 'str' thrown from the analyze link's navigate_dict, killing the chain and DLQing vCons.
• Root cause: navigate_dict was copy-pasted into 5 link modules with key in current traversal. When an intermediate step landed on a JSON-encoded body string (the spec-current shape after fix(vcon): decode body per spec on read, canonicalize on write #182), key in current did a substring check, passed, and the next current[key] raised on the string-indexed access.
• Replaced all 5 in-house copies with pydash.get (already a core dep — pydash>=7.0.7 in pyproject.toml). pydash.get handles every edge case we relied on (None input, missing key, non-dict mid-path, list-index-via-dot) and returns None cleanly instead of raising.
• Additionally wrapped source with Vcon.with_decoded_body(source) in analyze and analyze_and_label so a dotted text_location like body.paragraphs.transcript can still drill through a JSON-encoded body — check_and_tag and detect_engagement already do this; analyze/analyze_and_label had silently broken after fix(vcon): decode body per spec on read, canonicalize on write #182.
• Dropped 4 duplicate test_navigate_dict* blocks (testing in-house code that no longer exists). Kept one TestPydashGetContract::test_returns_none_when_midpath_is_non_dict in test_analyze.py that pins the CON-573 behaviour we depend on.
• Net diff: +52 / -131 across 9 files.

Side-effect (positive)

5 tests that were already failing on main with the same CON-573 TypeError now pass:

• conserver/tests/test_link_metrics.py::TestAnalyzeMetrics::test_success_records_analysis_time
• conserver/tests/test_link_metrics.py::TestAnalyzeMetrics::test_failure_increments_counter
• conserver/tests/test_link_metrics.py::TestAnalyzeAndLabelMetrics::test_success_records_labels_added
• conserver/tests/test_link_metrics.py::TestAnalyzeAndLabelMetrics::test_success_records_analysis_time
• conserver/tests/test_link_metrics.py::TestAnalyzeAndLabelMetrics::test_failure_increments_counter

Plus the 6 analyze_and_label/tests/test_analyze_and_label.py::test_run_* tests that were silently broken on main for the same reason.

The remaining test failures on this branch (TestCheckAndTagMetrics x3, TestOpenAITranscribeMetrics x2, groq_whisper::test_get_transcription) all pre-exist on main and are unrelated to this change.

Test plan

• pytest conserver/links/{analyze,analyze_and_label,check_and_tag,detect_engagement,analyze_vcon}/ — 38 passed, 10 skipped, 0 failures
• Reproduced the original TypeError against the pre-fix navigate_dict in a one-liner and confirmed the new code returns None instead
• Wider pytest conserver/ sweep — no new failures introduced
• Deploy to test env and confirm the original DLQ'd vCons re-process cleanly