Query-frontend: add -query-frontend.query-sharding-max-regexp-size-bytes

[pracucci]

What this PR does

Query sharding amplifies the number of partial queries executed by queriers, and so label matchers evaluated by ingesters and store-gateway. We've seen some edge cases where very complex regexp can spike the ingesters CPU significantly, up to the point of saturating the node CPU. Despite our efforts to optimise the regexp matcher (which in many cases are very effective), I would like to add an optional limit to query sharding, to skip it if the query contains a regexp longer than the configured limit.

Note to reviewers:

• The regexp length is not synonymous of complexity, but evaluating the actual computatinal complexity of a regexp may be computationally expensive itself, so just checking the length is a simple approximation I suggest to begin with.
• I removed TestQuerySharding_ShouldFallbackToDownstreamHandlerOnMappingFailure beacuse it expects to push down to queriers a query we can't parse, which is not correct (there's no point doing it). The test aims to
  ensure we push down to querier a query we fail to shard (but we can parse), but this can happen only in case of logic bugs in the query sharding, and so can't be reproduced unless mocking its internals. I tried to mock the internals to reproduce it, but the resulting code is pretty ugly, so I'm proposing to just get rid of that test.

Which issue(s) this PR fixes or relates to

N/A

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[pracucci]

Moving to draft because an integration test is failing in CI and I can't understand why. I'm investigating it.

[pstibrany]

PR lgtm. I can't see why integration tests would be failing.