test: xorb reconstruction cache warm-up benchmark and unit test

[XciD]

Summary

• Adds a unit test verifying warm_reconstruction_cache is triggered on open() and that the watch stays true for subsequent reads on the same handle (no re-waiting)
• Adds an EC2 integration benchmark (warm_cache_bench) measuring TTFB and sequential throughput for 5 consecutive opens of a 100 MB file

Background

The streaming path already used the xorb chunk disk cache correctly via FileReconstructor.with_chunk_cache(). This PR adds validation and numbers to prove it.

Benchmark results (m6i.2xlarge, /dev/shm cache)

Open#  TTFB (ms)  Total (s)  MB/s     cache
-----  ---------  ---------  -------  -----
    1      200.7       0.23    432.0  cold
    2        0.8       0.02   5741.3  warm
    3        0.7       0.01   6789.3  warm
    4        0.6       0.01   7050.2  warm
    5        0.6       0.01   7040.3  warm

cold TTFB 201 ms / 432 MB/s  →  warm TTFB 1 ms / 6.7 GB/s

The bench uses /dev/shm (tmpfs) on Linux so warm reads hit RAM. On EBS-backed /tmp (m6i default, ~125 MB/s baseline), warm reads would appear slower than cold S3 reads, making the cache benefit invisible.

Run

cargo test --release --test warm_cache_bench -- --nocapture

[github-actions]

POSIX Compliance (pjdfstest)

============================================================
  pjdfstest POSIX Compliance Results
------------------------------------------------------------
  Files: 130/130 passed    Tests: 832 total (0 subtests failed)
  Result: PASS
------------------------------------------------------------
  Category               Passed    Total   Status
  -------------------- -------- -------- --------
  chflags                     5        5       OK
  chmod                       8        8       OK
  chown                       6        6       OK
  ftruncate                  13       13       OK
  granular                    5        5       OK
  mkdir                       9        9       OK
  open                       19       19       OK
  posix_fallocate             1        1       OK
  rename                     10       10       OK
  rmdir                      11       11       OK
  symlink                    10       10       OK
  truncate                   13       13       OK
  unlink                     11       11       OK
  utimensat                   9        9       OK
============================================================

[github-actions]

Benchmark Results

============================================================
  Benchmark — 50MB
------------------------------------------------------------
  Metric                                 FUSE          NFS
  ------------------------------ ------------ ------------
  Sequential read                    193.1 MB/s     191.8 MB/s
  Sequential re-read                2252.8 MB/s    2276.7 MB/s
  Range read (1MB@25MB)                0.4 ms         0.2 ms
  Random reads (100x4KB avg)           0.0 ms         0.0 ms
  Sequential write (FUSE)           1033.3 MB/s
  Close latency (CAS+Hub)            0.132 s
  Write end-to-end                   277.9 MB/s
  Dedup write                       1037.1 MB/s
  Dedup close latency                0.069 s
  Dedup end-to-end                   428.0 MB/s
============================================================
============================================================
  Benchmark — 200MB
------------------------------------------------------------
  Metric                                 FUSE          NFS
  ------------------------------ ------------ ------------
  Sequential read                   1379.9 MB/s    1254.5 MB/s
  Sequential re-read                2209.6 MB/s    2379.9 MB/s
  Range read (1MB@25MB)                0.2 ms         0.2 ms
  Random reads (100x4KB avg)           0.0 ms         0.0 ms
  Sequential write (FUSE)           1009.0 MB/s
  Close latency (CAS+Hub)            1.665 s
  Write end-to-end                   107.3 MB/s
  Dedup write                       1078.9 MB/s
  Dedup close latency                0.142 s
  Dedup end-to-end                   611.5 MB/s
============================================================
============================================================
  Benchmark — 500MB
------------------------------------------------------------
  Metric                                 FUSE          NFS
  ------------------------------ ------------ ------------
  Sequential read                   1595.7 MB/s    1544.1 MB/s
  Sequential re-read                2170.5 MB/s    2422.8 MB/s
  Range read (1MB@25MB)                0.2 ms         0.2 ms
  Random reads (100x4KB avg)           0.0 ms         0.0 ms
  Sequential write (FUSE)            989.3 MB/s
  Close latency (CAS+Hub)            1.734 s
  Write end-to-end                   223.3 MB/s
  Dedup write                       1027.7 MB/s
  Dedup close latency                0.100 s
  Dedup end-to-end                   852.9 MB/s
============================================================
============================================================
  fio Benchmark Results
------------------------------------------------------------
  Job                        FUSE MB/s   NFS MB/s  FUSE IOPS   NFS IOPS
  ------------------------- ---------- ---------- ---------- ----------
  seq-read-100M                  505.0      286.5                      
  seq-reread-100M               1923.1     1408.5                      
  rand-read-4k-100M                0.4        0.4         98        114
  seq-read-5x10M                 574.7      312.5                      
  rand-read-10x1M                173.5       62.5      44428      16002
  Random Read Latency           FUSE avg      NFS avg
  ------------------------- ------------ ------------
  rand-read-4k-100M           10149.7 us    8756.3 us
  rand-read-10x1M                21.8 us      61.6 us
============================================================