release-25.4: tests: use admission.io.overload in admission-control/elastic-io

[blathers-crl]

Backport 1/1 commits from #155413 on behalf of @sumeerbhola.

---

The previously used sub-level metric was flawed, in that the IO overload score could stay low even at higher sub-level counts if L0 had very few bytes (which is a deliberate choice in admission control). So admission control would not throttle elastic work as aggressively as the test expected it to.

Running with this change, we don't exceed a score of 0.15, while the previously used sub-level count metric spikes higher. For example:

2025/10/14 22:13:15 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.100000
2025/10/14 22:13:25 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.100000

I251014 22:13:27.665296 868 util/admission/io_load_listener.go:780 ⋮ [T1,Vsystem,n1,s1] 2918  IO overload: compaction score 0.150 (131 ssts, 9 sub-levels), L0 growth 551 MiB (write 551 MiB (ignored 0 B) ingest 0 B (ignored 0 B)): requests 15985 (0 bypassed) with 505 MiB acc-write (0 B bypassed) + 0 B acc-ingest (0 B bypassed) + 551 MiB adjusted-LSM-writes + 4.2 GiB adjusted-disk-writes + write-model 1.09x+1 B (smoothed 1.08x+1 B) + l0-ingest-model 0.00x+0 B (smoothed 0.75x+1 B) + ingest-model 0.00x+0 B (smoothed 1.00x+1 B) + write-amp-model 7.87x+1 B (smoothed 8.01x+1 B) + at-admission-tokens 126 B, compacted 550 MiB [≈545 MiB], flushed 799 MiB [≈838 MiB] (mult 0.77); admitting 649 MiB (rate 43 MiB/s) (elastic 519 MiB rate 35 MiB/s) due to memtable flush (multiplier 0.775) (used total: 543 MiB elastic 541 MiB); write stalls 0; diskBandwidthLimiter (unlimited) (tokenUtilization 0.00, tokensUsed (elastic 0 B, snapshot 0 B, regular 0 B) tokens (write 0 B (prev 0 B), read 0 B (prev 0 B)), writeBW 0 B/s, readBW 0 B/s, provisioned 0 B/s)

2025/10/14 22:13:35 admission_control_elastic_io.go:105: admission_io_overload(store=1): 0.050000

I251014 22:13:42.666326 868 util/admission/io_load_listener.go:780 ⋮ [T1,Vsystem,n1,s1] 2926  IO overload: compaction score 0.050 (70 ssts, 5 sub-levels), L0 growth 498 MiB (write 498 MiB (ignored 0 B) ingest 0 B (ignored 0 B)): requests 15228 (0 bypassed) with 480 MiB acc-write (0 B bypassed) + 0 B acc-ingest (0 B bypassed) + 498 MiB adjusted-LSM-writes + 4.2 GiB adjusted-disk-writes + write-model 1.04x+1 B (smoothed 1.06x+1 B) + l0-ingest-model 0.00x+0 B (smoothed 0.75x+1 B) + ingest-model 0.00x+0 B (smoothed 1.00x+1 B) + write-amp-model 8.57x+1 B (smoothed 8.29x+1 B) + at-admission-tokens 153 B, compacted 498 MiB [≈522 MiB], flushed 883 MiB [≈860 MiB] (mult 0.77); admitting 667 MiB (rate 44 MiB/s) (elastic 533 MiB rate 36 MiB/s) due to memtable flush (multiplier 0.775) (used total: 519 MiB elastic 517 MiB); write stalls 0; diskBandwidthLimiter (unlimited) (tokenUtilization 0.00, tokensUsed (elastic 0 B, snapshot 0 B, regular 0 B) tokens (write 0 B (prev 0 B), read 0 B (prev 0 B)), writeBW 0 B/s, readBW 0 B/s, provisioned 0 B/s)

Fixes #148786
Fixes #156168
Fixes #156215

Epic: none

Release note: None

---

Release justification:

[blathers-crl]

Thanks for opening a backport.

Before merging, please confirm that the change does not break backwards compatibility and otherwise complies with the backport policy. Include a brief release justification in the PR description explaining why the backport is appropriate. All backports must be reviewed by the TL for the owning area. While the stricter LTS policy does not yet apply, please exercise judgment and consider gating non-critical changes behind a disabled-by-default feature flag when appropriate.

[cockroach-teamcity]

This change is