refactor(concurrent_delete): rename module + types to reflect N concurrent single-DELETE

[ldflags]

Closes #572.

Follow-up to issue #568 stage 6 (PR #569) and #570 (PR #571). Issue #568 stage 6 replaced the batch-XML DeleteObjects default path with rclone-style N concurrent single-DELETE workers. The module file src/batched_delete.rs and its types still carried the "batch" name from the retired implementation — there is no batching left, just an N-worker pool firing per-key DELETEs. This PR retitles everything to reflect the surviving design.

Identifier renames

Old	New
src/batched_delete.rs	src/concurrent_delete.rs
struct BatchedDeleter	ConcurrentDeleter
module batched_delete	concurrent_delete
WorkerConfig::batched_delete_config	WorkerConfig::concurrent_delete_config
MntrsFs::batched_deleter	MntrsFs::concurrent_deleter
tracing target mntrs::batched_delete	mntrs::concurrent_delete
log prefix "batched_delete: ..."	"concurrent_delete: ..."
error prefix "batched_delete error: ..."	"concurrent_delete error: ..."

Constants

• DEFAULT_BATCH_THRESHOLD removed from src/concurrent_delete.rs.
• Two new consts added at the top of src/lib.rs (above UnlinkBurstState):
  • DEFAULT_UNLINK_BURST_WINDOW_MS: u64 = 100
  • pub(crate) DEFAULT_UNLINK_BURST_THRESHOLD: u32 = 32
• UnlinkBurstState::from_env now reads these two consts instead of the old batched_delete::DEFAULT_BATCH_THRESHOLD reference.

Scripts / bench

• bench/unlink_ab.sh — strip dead MNTRS_BATCH_FLUSH_DELAY_MS=10 export; retarget tracing target; rewrite the batch-XML metrics block (counts batch_size=N / multi-key lines) to count single-key DELETE / retry / fail events from the N concurrent single-DELETE worker pool.
• scripts/micro_bench_rm.sh — header comment updated to "S3 deleter paths"; strip dead MNTRS_BATCH_FLUSH_DELAY_MS=10, MNTRS_BATCH_THRESHOLD=0, MNTRS_BATCH_SIZE=20, MNTRS_BATCH_FAST_FLUSH_THRESHOLD=8 exports; retarget tracing target; drop the now-obsolete 3-mode comparison block ("batched+fast-flush" mode is gone).
• bench/run_all.sh — rewrite the knob doc block from the retired four-knob shape (MNTRS_BATCH_FLUSH_DELAY_MS / MNTRS_BATCH_THRESHOLD / MNTRS_BATCH_SIZE / MNTRS_BATCH_FAST_FLUSH_THRESHOLD / MNTRS_BATCH_PROFILE) to the surviving two knobs (MNTRS_BURST_WINDOW_MS=200, MNTRS_BURST_THRESHOLD=4); strip the dead env exports from the mount command; retarget tracing target.

Out of scope (preserved as-is)

• MNTRS_BATCH_MNT / MNTRS_BATCH_DAEMON_LOG shell vars: mount path and log path, not runtime knobs.
• delete_mode = "concurrent-single" literal: already aligned in PR batched_delete: N concurrent single-DELETE workers (issue #568) #569.
• bench/delete_batch_poc.sh: standalone PoC for the retired DeleteObjects XML path, not invoked from the bench flow.
• Historical references to "batched_delete" in explanatory comments (Issue design RFC: batched_delete target system (N-consumer × profile × burst observer) #562 stage 0 workload rationale, the "Removed 2026-08-03" mkdir-deep test record, the dead-knobs archive comment in src/concurrent_delete.rs) are preserved verbatim — they document the design history that this rename is a follow-up to.

Verification

• cargo fmt --all -- --check — green
• cargo clippy --workspace --all-targets -- -D warnings — green
• cargo build (debug) — green
• cargo test --workspace --lib — 281 passed, 0 failed
• bash -n bench/run_all.sh bench/unlink_ab.sh scripts/micro_bench_rm.sh — green
• git grep -E "batched_delete|BatchedDeleter|batched_deleter" in src/, bench/, scripts/ — 0 hits in active code; remaining matches are historical doc-comments and the standalone delete_batch_poc.sh PoC (out of scope).