Add ESR API receipts and typed public archive samples

[kvkenyon]

Public archive failures previously lacked exact transfer receipts, and the separate ESR REST service had no adapter. Add direct-import ESR raw access and archive clients with bounded response receipts, strict content types, and explicit failure boundaries. Add typed samples for two source-verified nested CSV formats.

• Installed retrievals: DME ZIP 22,983 bytes; August 2026 capacity bundle 34,664 bytes; retired NP4-179-CD archive 1,148 bytes. The installed wheel decoded four typed rows from each exact cached file.
• ESR current returns HTTP 401 with an invalid-subscription-key message while the same token succeeds on Public Reports. ESR row schemas remain unknown. The oldest listed DAM bundle still returns HTTP 400; another negative-ID bundle succeeds with the documented POST/docIds shape.
• Public BINARY documents and retired history remain supported or outstanding file access. No full-history claim, restricted access, bulk download, or shared registration change.

Validation: 1,318 tests passed, one skipped; 59 dedicated tests; generation, lint, formatting, wheel build, installed PEP 561 proof, and direct-module mypy proof passed. Credential exclusion and exact installed module bytes were checked.

Stacked on PR #4's remote head, feat/public-metadata-expansion, from immutable commit 11e20c84b11a38186ef1fe3e05a615de2cb8571d. All changes use new assigned paths. See docs/esr-archive-task-note.md for commands, receipts, precise gaps, and the import/registry handoff. Do not merge as part of this task.