Use sentry performance profiling for swh-web in staging

Hey ops, I pushed swh-code@3.7.0 that allows us to use sentry performance profiling. It would be great to have it deployed on staging at some point.

By default profiling is disabled, to enable it a dedicated env var SWH_SENTRY_TRACES_SAMPLE_RATE is needed. Would it be possible to try it for swh-web in staging with SWH_SENTRY_TRACES_SAMPLE_RATE=0.5 (i.e. 50% of the transactions will be profiled)?

As @olasd suggested in the MR it might be a good idea to keep an eye on sentry.swh.org's metrics to see how this impacts its workload and tweak SWH_SENTRY_TRACES_SAMPLE_RATE value to see if/when it breaks.

For it to be most useful it really needs to be enabled for all backend services (at least in the read-only path), rather than just swh-web, doesn't it?

Trace propagation sure looks nice, but for now even without a detailed view of what's happening during API calls to another swh service I think we'll get some interesting data.

I thought that initially we could enable this little by little as I feared it could 1) impact our infrastructure too much 2) be overwhelming to analyze too much data at once.

But if you think we should activate that on the entire stack, it's okay with me.

I think setting SWH_SENTRY_TRACES_SAMPLE_RATE=0 on backend services should keep the sample rate fully constrained to whatever swh-web uses.

FWIW I poked at the image building jenkins job (which failed due to a race condition in parallel git pushes...) and swh/infra/ci-cd/swh-charts!533 (merged) eventually happened.

Hey ops, I pushed swh-code@3.7.0 that allows us to use sentry performance profiling. It would be great to have it deployed on staging at some point.

The core 3.7 has been deployed along with #5546 (closed) btw. But no setup has been done just yet.