core: identify failed spclient routes safely

[dereksix]

What changed

• Emit one warning after an spclient request exhausts its retry strategy.
• Include only the HTTP method, path-only endpoint, attempt count, and broad ErrorKind.
• Strip query strings and omit headers, bodies, tokens, and the underlying error detail.
• Add a regression test for endpoint sanitization.

Why

Previously the final request error was returned without enough route context to distinguish failures in storage resolution, context/autoplay, playlists, or Connect state. That made intermittent Spotify failures difficult to diagnose without enabling much noisier debug logging.

This is observability-only; it does not change request construction, retry behavior, access-point rotation, or error propagation.

Validation

• cargo fmt --all -- --check
• cargo test -p librespot-core (17 unit tests, Connect integration test, and doctests passed)
• cargo clippy -p librespot-core --all-targets -- -D warnings