DynamicFootprint: confluence example and tests/test_2

[sstruzik]

Companion to OasisLMF/OasisLMF#2219 (issue OasisLMF/OasisLMF#2218). Merge OasisLMF#2219 first: this model's lookup now uses the join_rows step, which only exists there, so this PR's automatic CI run (oasislmf from main) fails until it lands. To test both together beforehand, run the model-tests workflow by hand with oasislmf_branch = feature/confluence-section-merge.

What changes in DynamicFootprint

A confluence example: areaperils lying in several sections, which one event can drive at once.

• Sections 3001-3003 overlap on 18 existing areaperils of the grid (columns 0, 1 and 9 of rows 2-7). Each row is one combination of sections (S1, S1+S2, S2, S1+S3, S1+S2+S3, S2+S3), each column one hazard uncertainty variant: none, the same number of realisations everywhere, or different numbers between sections and between return periods.
• Three new hazard case partitions and new event definition rows for events 1-8. The existing partitions, the other events' rows and section 2143's polygon are untouched.
• The cells avoid section 2143's polygon, whose edges run along the grid's diagonals (a point on them falls in or out on the seventh decimal), and tests/test_1's locations. test_1's portfolio only uses section 2143, so it never loads the new sections.
• The hazard case has no probability column, like the rest of the model: realisations are written as repeated rows (equal weights), e.g. .25/.5/.25 as one value once, one twice, one once.
• Intensities stay within intensity bins 1-29, which the model's (PiWind) vulnerability covers.
• Lookup: a join_sections step (type join_rows) after the section rtree gives a location in several sections one keys value (3001;3002).
• tests/test_2: the 18 locations, one loss line each, default section merge (max).
• make_confluence_data.py adds the example to the model data, and only it; show_confluence.py prints the footprint OasisLMF builds at those areaperils, without a model run.

Verification (with OasisLMF#2219 installed)

• test_1: output byte-identical to its existing expected results.
• test_2: every location gets exactly its sections in the keys; the footprint matches hand-computed values (e.g. event 1: S1+S3 gives 13 at rp 50, every region containing S2 gives 15 at rp 200); 126 event x location losses, all positive.
• Both tests pass --check-results.

🤖 Generated with Claude Code