OOM Builds get retried infinitely

llvm-toolchain-17 is OOMing and was retrying infinitely.

I think this is because the aborted task never results in a recorded failed state: https://salsa.debian.org/freexian-team/debusine/-/blob/67791b88c150b075aba0c487317c64760cb5a6c4/debusine/worker/_worker.py#L581-584

This dates back to !142 (merged).

It's nice to automatically retry a build after the build was killed by an admin, or something like that. But within limits.

The simple fix here is to record the failure, instead of letting the build retry. Maybe a better option is to implement a retry limit (by counting the number of logs?)

Stopped it manually with:

debusine=> UPDATE db_workrequest SET status='failed', result='error' WHERE id=3400;
UPDATE 1