Use hyper 1.0 and tonic 0.12 in storage broker

[arpad-m]

Fixes #9231 .

Upgrade hyper to 1.4.0 and use hyper 1.4 instead of 0.14 in the storage broker, together with tonic 0.12. The two upgrades go hand in hand.

Thanks to the broker being independent from other components, we can upgrade its hyper version without touching the other components, which makes things easier.

[github-actions]

5084 tests run: 4898 passed, 0 failed, 186 skipped (full report)

---

Flaky tests (1)

Postgres 14

• test_timeline_copy[100000]: release-x86-64

Code coverage* (full report)

• functions: 31.3% (7487 of 23882 functions)
• lines: 49.5% (60108 of 121323 lines)

* collected from Rust tests only

---

_{The comment gets automatically updated with the latest test results
3109c4b at 2024-10-02T21:17:51.259Z :recycle:}

[koivunej]

I think this is fine, but I would add a comment or assertion/expect/unreachable why we don't expect serve to fail anymore.

use of http_body_util::Either can be done later.

[koivunej]

Hmm I think we should change "Fixes" to "Cc" or something similar in PR description, because we will still have hyper pre-1.0, or open another issue for that?

[koivunej]

hakari doesn't like the toml: https://github.com/neondatabase/neon/actions/runs/11138756565/job/30954289891?pr=9234

[arpad-m]

I think we should change "Fixes" to "Cc" or something similar in PR description, because we will still have hyper pre-1.0, or open another issue for that?

IMO the issue is mainly about updating tonic, and the parts about updating hyper are to update hyper in enough places that a tonic update is possible. This is accomplished by this PR.

[koivunej]

I think the current impl does not serve more than one connection at a time, but the old one did. I think we should had had some tests fail for this.

[koivunej]

I am still unsure why not a lot of test failures with regress tests if storage_broker only served a single connection (pageserver or safekeeper).