feat(computing-unit): add per-computing-unit dataset mount API

[aicam]

What changes were proposed in this PR?

Add a backend API to mount, list, and unmount dataset versions on a specific computing unit, independent of any workflow. The computing-unit service acts as an authenticated, stateless proxy to that CU's node texera-mounter — mount state lives only on the mounter (the kernel mount table), so nothing is persisted here.

• Node mounter (bin/mounter/mounter.py) — GET /mounts?cuid= lists a CU's current mounts (from /proc/mounts); DELETE /mount unmounts a single dataset (distinct from the pod-death reaper).
• Computing-unit service (ComputingUnitManagingResource) — GET/POST/DELETE /computing-unit/{cuid}/mounts. POST resolves /ownerEmail/datasetName/versionName → repository:commitHash, finds the CU pod's node IP, and forwards to that node's mounter with a freshly minted per-user JWT (authorization stays in file-service). A small MounterClient speaks HTTP to the mounter.
• Reverse resolution (FileResolver) — repository:commitHash → /ownerEmail/datasetName/versionName, so listed mounts show a human-readable path.

Purely additive — no existing endpoint, schema, or runtime path changes. This is the control-plane foundation for the "Mount datasets into computing unit" UI (follow-up).

Any related issues, documentation, discussions?

Implements apache#6862 · part of apache#6606. Stacks on apache#6608 (mount infrastructure).

How was this PR tested?

Compiled (ComputingUnitManagingService/compile) and validated end-to-end on a single-node minikube: created an 8Gi Kubernetes computing unit, POST /computing-unit/{cuid}/mounts returned 200 with the host mount path, GET /mounts listed it with the reverse-resolved dataset path, and DELETE unmounted it (empty list after). The mounted dataset was then read bit-exact by a Python UDF (a ~2 GB sharded PyTorch model, torch.load from the mounted path).

Was this PR authored or co-authored using generative AI tooling?

Generated-by: Claude Opus 4.8

[github-actions]

Automated Reviewer Suggestions

Based on the git blame history of the changed files, we recommend the following reviewers:

• Contributors with relevant context: @Ma77Ball
  You can notify them by mentioning @Ma77Ball in a comment.

[github-actions]

⚠️ Benchmark changes need a look

🟢 0 better · 🔴 7 worse · ⚪ 8 noise (<±5%) · 0 without baseline

Compared against main 9602949 benchmarked on this same runner, so the delta is largely free of cross-runner hardware noise. The "7d avg" column still reflects the gh-pages dashboard. Treat <±5% as noise unless repeated.

Dashboard · Run

	config	throughput	MB/s	latency	max Δ latest / 7d
🔴	bs=10 sw=10 sl=64	365	0.223	26,610/45,550/45,550 us	🔴 +20.4% / ⚪ within ±5%
🔴	bs=100 sw=10 sl=64	802	0.489	123,509/148,981/148,981 us	🔴 +5.5% / ⚪ within ±5%
⚪	bs=1000 sw=10 sl=64	911	0.556	1,094,336/1,155,758/1,155,758 us	⚪ within ±5% / ⚪ within ±5%

Baseline details

Latest main 9602949 from same runner

config	metric	PR	latest main	7d avg	Δ latest	Δ 7d
bs=10 sw=10 sl=64	throughput	365 tuples/sec	397 tuples/sec	n/a	-8.1%	n/a
bs=10 sw=10 sl=64	MB/s	0.223 MB/s	0.243 MB/s	n/a	-8.2%	n/a
bs=10 sw=10 sl=64	p50	26,610 us	22,431 us	n/a	+18.6%	n/a
bs=10 sw=10 sl=64	p95	45,550 us	37,839 us	n/a	+20.4%	n/a
bs=10 sw=10 sl=64	p99	45,550 us	37,839 us	n/a	+20.4%	n/a
bs=100 sw=10 sl=64	throughput	802 tuples/sec	808 tuples/sec	n/a	-0.7%	n/a
bs=100 sw=10 sl=64	MB/s	0.489 MB/s	0.493 MB/s	n/a	-0.8%	n/a
bs=100 sw=10 sl=64	p50	123,509 us	121,058 us	n/a	+2.0%	n/a
bs=100 sw=10 sl=64	p95	148,981 us	141,180 us	n/a	+5.5%	n/a
bs=100 sw=10 sl=64	p99	148,981 us	141,180 us	n/a	+5.5%	n/a
bs=1000 sw=10 sl=64	throughput	911 tuples/sec	921 tuples/sec	n/a	-1.1%	n/a
bs=1000 sw=10 sl=64	MB/s	0.556 MB/s	0.562 MB/s	n/a	-1.1%	n/a
bs=1000 sw=10 sl=64	p50	1,094,336 us	1,077,196 us	n/a	+1.6%	n/a
bs=1000 sw=10 sl=64	p95	1,155,758 us	1,140,117 us	n/a	+1.4%	n/a
bs=1000 sw=10 sl=64	p99	1,155,758 us	1,140,117 us	n/a	+1.4%	n/a

Raw CSV

config_idx,batch_size,schema_width,string_len,num_batches,total_ms,total_tuples,total_bytes,tuples_per_sec,mb_per_sec,lat_p50_us,lat_p95_us,lat_p99_us
0,10,10,64,20,547.65,200,128000,365,0.223,26610.06,45549.56,45549.56
1,100,10,64,20,2494.54,2000,1280000,802,0.489,123508.59,148981.40,148981.40
2,1000,10,64,20,21961.21,20000,12800000,911,0.556,1094335.55,1155758.17,1155758.17

[aicam]

Superseded — this feature is now being contributed directly to apache/texera:main as two balanced PRs: apache#6866 (out-of-pod mount infrastructure) and apache#6867 (platform mount integration), tracked in apache#6606. Closing this fork-staging PR.