feat(esm_catalog): PR-A1e — namelist extension

[siligam]

Part of the esm_catalog decomposition, per the PR-A1a foundation plan. Follows #1504 (A1b core), branches from release.

Adds src/esm_catalog/namelist.py:

• add_namelist_extension(collection, namelists) — collection-level nml:files, nml:groups, nml:parameters (flattened "group:key" → scalar, for CQL2 filtering). Drops nml:raw from the pr-a1a version per the plan — storing the full raw namelist as JSON on every collection could be megabytes for complex setups.
• add_namelist_item_extension(item, ctx) — item-level nml:{component}:{group}:{key} entries across all components, read from ctx.namelists_by_component (kept as-is per the plan; performs no scanning).

Wired into make_collection (using the collection's own component's entry from ctx.namelists_by_component) and make_item.

Also adds configs/stac-extensions/namelist/v1.0.0/schema.json (custom extension — no external community schema exists) and tests/test_esm_catalog/test_stac_namelist_ext.py.

Test plan

• pytest tests/test_esm_catalog — 51 passed
• New tests cover: collection no-op/flattening/files+groups/nml:raw dropped/nested-dict+long-list skipping/idempotent URL append, item no-op/multi-component coverage, and schema validation (via jsonschema) for both Collection and Item shapes

[siligam]

@mandresm heads-up: per our Webex split this one was yours — I went ahead and implemented it anyway while I had momentum finishing the A1 plan (A1c/A1f landed as #1511/#1513 the same day, and this one follows the plan's spec directly: nml:raw dropped, files/groups/parameters kept, item-level function unchanged). If you'd already started on it or want it done differently, happy to close this in favour of yours — otherwise it's ready for your review.