Document settings, repeated runs and scenario comparison

[TimoDiepers]

Merge last. This sits on top of #219, #222 and #223, because it documents features from all three — it merges #222 into #223's branch so the docs can reference create_missing and compare() in the same breath. Until those land, the diff below includes their commits; the two commits belonging to this PR are the merge and the docs commit at the tip.

Strategy: one spine, one topic page — not two tracks

run() calls the same four steps the Walkthrough teaches, in the same order. Documenting it as a parallel "track" would imply a choice that doesn't exist, duplicate every concept, and leave newcomers picking between two ways to do one thing — while the procedural API stays necessary anyway (you inspect tlca.timeline between steps, re-run one step with different arguments).

So this follows the shape scikit-learn uses for Pipeline/GridSearchCV: fit/predict stay the spine, and "run many of them" is its own topic page that assumes you already know the spine. Same pattern as requests (Quickstart vs. Session in Advanced) and pytest (plain asserts, then fixtures).

The Walkthrough is untouched, except one closing pointer in Step 4.

What's here

New page — Repeated Runs & Scenario Comparison (content/getting_started/scenarios.md), placed after the Walkthrough:

• one calculation as one TimexLCASettings, via from_settings().run()
• re-running with per-call overrides, with a table of what each kind of change reuses vs. rebuilds — including why a changed background is refused, and where to go instead
• compare() and ComparisonResult.summary, plus keep_objects / on_error
• where the scenario databases come from — linking to create_missing in the Quick Start rather than restating it

Supporting edits, deliberately small:

• Step 2 claimed "comparing scenarios is the same script in a loop over filters" — stale now that compare() exists; it points there instead
• Quick Start gains a six-line "steps 5–7 in one call" snippet beside the procedural example, plus a run()/compare() reference table
• the API page introduces the two new classes (mkdocstrings already picks them up from the module)

Verification

Every snippet on the new page was executed against a throwaway project before being written down — from_settings/run, the overrides, the refused-background ValueError, and compare()'s summary columns — so the shapes and claims match the code rather than my memory of it. Internal links and cross-page anchors checked programmatically. Full suite across the merged branches: 440 passing.

[codecov]

Codecov Report

❌ Patch coverage is 93.15068% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 84.50%. Comparing base (1bb7fb4) to head (dae42df).
⚠️ Report is 30 commits behind head on main.

Files with missing lines	Patch %	Lines
bw_timex/timex_lca.py	93.15%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #224      +/-   ##
==========================================
+ Coverage   83.37%   84.50%   +1.13%     
==========================================
  Files          14       15       +1     
  Lines        3013     3428     +415     
==========================================
+ Hits         2512     2897     +385     
- Misses        501      531      +30     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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