ci: bump MSan fuzz timeout from 150 to 180 minutes #35526

The MSan fuzz job is cutting it a bit close on forks (with no cache hits):

Examples from Sjors/bitcoin:

• PR 116 timed out after 2h30m, and would have finished ~5 mins later: https://github.com/Sjors/bitcoin/actions/runs/27465468297/job/81187008305?pr=116
• PR 117 passed in 2h24m47s: https://github.com/Sjors/bitcoin/actions/runs/27465461797/job/81186956747?pr=117
• PR 118 passed in 2h18m22s: https://github.com/Sjors/bitcoin/actions/runs/27465761345/job/81187860567?pr=118

Bumping the timeout to 180 minutes should make these timeouts sufficiently rare, and not make much of a difference here.

<!--e57a25ab6845829454e8d69fc972939a-->

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

<!--006a51241073e994b41acfe9ec718e94-->

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/35526.

<!--021abf342d371248e50ceaed478a90ca-->

Reviews

See the guideline for information on the review process.

If your review is incorrectly listed, please copy-paste <code>&lt;!--meta-tag:bot-skip--&gt;</code> into the comment that the bot should ignore.

<!--5faf32d7da4f0f540f40219e4f7537a3-->

This was last bumped in #33626 for the same reason, so either the tests have grown or there's shrinkflation on the runners?

Update: my LLM friend says the corpus grew by 25% since then, so that's probably the reason.

lgtm ACK 17353f9d977db7bb86ad04da03e2f7c6a5309e45

The other fuzz task has a timeout of 240min, so 180min here seems fine, or even just 240 as well.

Update: my LLM friend says the corpus grew by 25% since then, so that's probably the reason.

Yeah, I'd think this is mostly the accumulated fuzz inputs over time. After https://github.com/bitcoin-core/qa-assets/issues/265, the extremely unlikely worst case could be 3x overhead, so in theory it could be bumped more, but in practise this may not be needed.

lgtm either way

ACK 17353f9d977db7bb86ad04da03e2f7c6a5309e45

Yeah, I'd think this is mostly the accumulated fuzz inputs over time. After bitcoin-core/qa-assets#265, the extremely unlikely worst case could be 3x overhead, so in theory it could be bumped more, but in practise this may not be needed.

YOu can see more inputs being added periodically on this runtime execution chart: https://willcl-ark.github.io/bitcoin-core-ci-stats/graph/execution/#MSan,%20fuzz

It says we usually prune inputs around release time over in that repo, but perhaps we could do a run sooner, as recent merged have slowed down CI a fair bit?