parity(#252): make the 1 MB max file size configurable (CSP_MAX_FILE_BYTES) and warn about skipped files

[amondnet]

Summary

Upstream semble #252 (9bb38fd) made the 1 MB per-file indexing limit configurable via SEMBLE_MAX_FILE_BYTES and warns about files skipped for size (previously silent, leaving unexplained gaps in results — upstream #250). csp hardcodes the limit and skips silently.

Upstream

• index/files.py::get_max_file_bytes(): reads the env var; malformed or non-positive values warn and fall back to the 1 MB default.
• index/create.py: collects FileStatus.TOO_LARGE paths and logs one warning naming up to 5 of them (... raise SEMBLE_MAX_FILE_BYTES to include them).
• cli.py::_configure_cli_logging(): idempotent stderr handler so the warning surfaces in CLI runs.
• README: documents the limit and the env var.

csp status

• crates/csp/src/indexing/create.rs: pub const MAX_FILE_BYTES: u64 = 1_000_000; and if size > MAX_FILE_BYTES { continue; } — hardcoded, silent.
• csp has no CSP_* env overrides at all (upstream also has SEMBLE_CACHE_LOCATION, SEMBLE_MODEL_NAME, SEMBLE_CLONE_TIMEOUT); scope this issue to the file-size limit, note the others as a possible follow-up.

Tasks

• CSP_MAX_FILE_BYTES env override with the same fallback/warn semantics.
• Collect skipped-for-size paths during create_index_from_path and emit one stderr warning (first 5 paths + count).
• README.md + README.ko.md: document the limit and the env var.
• Tests: env override honored; invalid value falls back; warning emitted.

Refs

• Upstream: Make max file size configurable via SEMBLE_MAX_FILE_BYTES and warn on skipped files MinishLab/semble#252 (9bb38fd), issue Files larger than 1 MB are silently skipped during indexing - no warning, no way to raise the limit MinishLab/semble#250
• Sync baseline check: 949559b..6bbbd32, 2026-09-03