fix: purge restored Burst cache in Cloud Build PreExport

[aixaCode]

Pull Request Description

What does this PR change?

Windows Cloud Builds have been failing intermittently (24/55 runs on Jun 10) with:

Burst compiler failed running
Error: Failed to find entry-points:
System.Exception: Error while loading assembly references for AppArgs
  ---> System.IO.EndOfStreamException: Unable to read beyond the end of the stream.
  at Burst.Compiler.IL.Hashing.Storage.ILHashingResultRuntime.GetHeader(String hashCachePath)

The Library cache restored by Unity Cloud Build contains truncated Burst hash-cache files — most likely from builds cancelled mid-cache-write (our cancel-in-progress cancels in-flight Cloud Builds on every push). Burst crashes on the corrupt file instead of treating it as a cache miss, and the poisoned target then fails every incremental build until someone runs a clean build. New PR targets can also inherit a poisoned cache from the template target via buildTargetCopyCache. Evidence it's not code: the same commit succeeded and failed 24s apart (runs 27354802918 / 27354828088), and no package/Unity/asmdef changes landed when the spike started.

This PR deletes Library/BurstCache in CloudBuild.PreExport() — which runs on the build worker after cache restore, before the Burst AOT step — so the cache always regenerates. Costs a few minutes of Burst recompile per build; a poisoned target costs a failed ~30 min build plus human retries. Non-fatal on error (logs a warning and proceeds). No Cloud Build config changes needed: PreExport is already wired in the template targets all PR targets are cloned from.

Draft for testing: verifying on this PR's own builds that [BURST]: purged restored Library/BurstCache appears in the log and the EndOfStreamException failures stop, plus measuring the build-time cost before merging.

Test Instructions

Steps (standard run): N/A — build infrastructure change only.

Expected result: N/A

Steps (fresh account): N/A

Expected result: N/A

Automation (if applicable): N/A

Prerequisites

• None — the change takes effect on any Cloud Build of this branch

Test Steps

1. Let this PR's Unity Cloud Build run.
2. Check the build log for [BURST]: purged restored Library/BurstCache (or the "nothing to purge" line on a clean target).
3. Compare Unity ran in N seconds against recent dev builds to measure the Burst recompile cost.
4. Re-run the build a few times — no EndOfStreamException / "Failed to find entry-points" failures should occur.

Additional Testing Notes

• If failures persist with the purge log present, the AOT compiler reads a cache outside Library/BurstCache (e.g. under Library/Bee) and the purge path needs widening.
• Follow-up worth checking: whether a Burst patch newer than 1.8.28 handles corrupt hash caches as cache misses.

Quality Checklist

• Changes have been tested locally
• Documentation has been updated (if required)
• Performance impact has been considered
• For SDK features: Test scene is included

[github-actions]

Windows and Mac build successful in Unity Cloud! You can find a link to the downloadable artifact below.

Name	Link
Commit	8062aac
Logs	https://github.com/decentraland/unity-explorer/actions/runs/27370038228
Download Windows	https://github.com/decentraland/unity-explorer/suites/73606781497/artifacts/7574974963
Download Windows S3	https://explorer-artifacts.decentraland.org/@dcl/unity-explorer/branch/fix/purge-burst-cache-cloud-build/pr-22966-8062aac/Decentraland_windows64.zip
Download Mac	https://github.com/decentraland/unity-explorer/suites/73606781497/artifacts/7574793443
Download Mac S3	https://explorer-artifacts.decentraland.org/@dcl/unity-explorer/branch/fix/purge-burst-cache-cloud-build/pr-22966-8062aac/Decentraland_macos.zip
Built on	2026-06-11T19:41:41Z

[m3taphysics]

PR #8966, run #27372815381

Builds: Windows change, Windows baseline, macOS change, macOS baseline

How to read this table

• Each build is measured 3 times. The values are the median, and (min–max) is the lowest and highest of those runs — a wide range means the metric is noisy and small differences are not trustworthy.
• Δ is Change minus Baseline (a negative Δ means Change is faster).
• 🟢 faster / 🔴 slower — a real difference: larger than both 3% and the run-to-run range.
• ⚪ within noise — the difference is smaller than how much the build varies between its own runs, so it cannot be told apart from random variation. Treat it as no change.
• Exceptions logged — exceptions found in the run logs; more than the baseline is flagged 🔴 even when frame times look fine. The Exception breakdown under each table groups them by the explorer's report category and exception type.
• A run that logged unusually many exceptions (at least 10 and 5× the median of its build's runs — e.g. a service was down during it) is excluded from all numbers and called out under the table.

Framework 13 i7

Metric	Baseline	Change	Δ	Result
Samples	2701 (×3)	2701 (×3)
CPU average	33.3 ms	33.3 ms	-0.0 ms	⚪ within noise
CPU 1% worst	33.8 ms (33.7–34.0)	33.6 ms (33.5–33.6)	-0.2 ms	⚪ within noise
CPU 0.1% worst	37.1 ms (36.0–37.2)	35.2 ms (34.4–36.2)	-1.9 ms	🟢 5% faster
GPU average	7.5 ms (7.5–7.6)	7.5 ms (7.5–7.8)	0.0 ms	⚪ within noise
GPU 1% worst	19.7 ms (19.2–20.0)	19.0 ms (18.9–20.9)	-0.7 ms	⚪ within noise
GPU 0.1% worst	24.8 ms (24.1–26.8)	23.9 ms (22.9–30.9)	-0.9 ms	⚪ within noise
Exceptions logged	6	6	0	⚪ none new

Exception breakdown

Exception	Baseline	Change
[ENGINE] NullReferenceException	6	6