docs(tutorials): add the FASTA-to-scikit-learn upstream bridge recipe

[breimanntools]

The upstream adapters exist as code but none was documented as an end-to-end bridge. tutorials/tutorial8_upstream_bridge.ipynb takes a user from another tool's output to a fitted model.

FASTA in → read_fasta → get_df_parts → a genuine Pipeline([SequenceFeatureTransformer, StandardScaler, RandomForestClassifier]) under cross_val_score → get_feature_names_out showing the CPP feature ids survive into RF importances → a second unlabeled FASTA through transform / predict.

Zero library code touched — 3 files: the notebook, a gallery tile, and the tutorials.rst entry under a new Interoperability section.

Two judgement calls

1. It uses SequenceFeatureTransformer, not aa_composition. Codex proposed the latter, which is a trivial pipeline that dodges the actual gap: the leak-free CPP-selection-inside-CV pipeline that today exists only in tests/integration/test_sklearn_pipeline_compat.py. The recipe is lifted out of that test — which is what the issue asks for, since it notes the only runnable pipeline lives in a test file.
2. It exports the bundled DOM_GSEC (40 records) to a real FASTA rather than hand-writing six mock records, so the 5-fold CV is meaningful rather than decorative. Still offline, still no new dependency.

The one-bridge-versus-every-bridge contradiction in the issue is resolved in favour of one, matching its own Scope / non-goals line — and resolved inside the docs, so the next reader does not rediscover it: the notebook and the new section both say only the FASTA hand-off is covered and point at the embeddings/AlphaFold tutorial for the others.

Verification

Docs build before and after: 0 error · 167 critical both times, critical did not move, and the new page was confirmed built (41 KB RST, 85 KB HTML, 5 tables, the figure rendered). pytest --nbmake on the notebook: 1 passed in 64s — re-run independently here at 98s. 341 api_tests pass.

One open item

tutorials/ is not executed in CI at all — the notebook gate covers an examples/ subset plus all of protocols/. So this notebook has no blocking execution gate. Closing that is a one-line workflow addition (~65s to that job), left for you since .github/workflows/* is CONFIRM-FIRST.

Refs #555.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.39%. Comparing base (3c11c90) to head (0ddb4d0).
⚠️ Report is 10 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #595   +/-   ##
=======================================
  Coverage   95.39%   95.39%           
=======================================
  Files         222      222           
  Lines       23387    23387           
  Branches     4073     4073           
=======================================
  Hits        22309    22309           
  Misses        631      631           
  Partials      447      447           

see 2 files with indirect coverage changes

Components	Coverage Δ
cpp_core	95.97% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.