Don't fail fast CI workflow (#442)

With flaky tests or Docker rate limit, it's very annoying to have a
job's failure cancel another job. Instead, let all jobs run to
completion, this way we can restart only the failed one.

.github/workflows/ci.yaml

@@ -24,6 +24,7 @@ jobs:
   build:
     name: CI
     strategy:
+      fail-fast: false
       matrix:
         archBox:
         - { arch: amd64, vmArch: x64 }