GHA cache backend connection resets may cause build failure

[jeroenvisser101]

Since there is no retry mechanism, we've seen some of our builds fail because of intermittent connection issues while attempting to store the cache. It may be better to retry this operation, or to warn of a failure to store the cache but not to crash the build.

I've attached the logs with identifiers removed.

We deploy 38 different sites using a matrix, and limit parallelism to 20 to avoid overloading anything, but this still happens just about every build.

#32 exporting cache
#32 preparing build cache for export
#32 preparing build cache for export 4.8s done
#32 ERROR: Get "https://artifactcache.actions.githubusercontent.com/[...]/_apis/artifactcache/cache?keys=buildkit-blob-1-sha256%3A[...]&version=[...]": read tcp 172.17.0.2:34032->13.107.42.16:443: read: connection reset by peer
------
 > exporting cache:
------
error: failed to solve: Get "https://artifactcache.actions.githubusercontent.com/[...]/_apis/artifactcache/cache?keys=buildkit-blob-1-sha256%3A[...]&version=[...]": read tcp 172.17.0.2:34032->13.107.42.16:443: read: connection reset by peer
Error: buildx failed with: error: failed to solve: Get "https://artifactcache.actions.githubusercontent.com/[...]/_apis/artifactcache/cache?keys=buildkit-blob-1-sha256%3A[...]&version=[...]": read tcp 172.17.0.2:34032->13.107.42.16:443: read: connection reset by peer

[crazy-max]

#681 (comment)