chore: Improve networking code pt 1

[ajbt200128]

This PR is part 1 of 2 of improvements I plan to make to the networking library. The next PR will introduce a custom client that will (optionally) use proxies when establishing connections directly, instead of the hacky way we're doing rn that abuses Cohttp.

Changes:

• remove lwt platform functor
  • this was ugly, and we use the async functions pretty much everywhere already, so let's just commit
  • consumers of http_helpers can use lwt_platform.run if they need sync
• Bump cohttp to 6.0.0 beta 2
  • @akuhlens has a PR that will come out in the release version of the library, and there's been a bunch of improvements anyways, so let's just use it
  • The proxy changes I'm going to make I want to upstream, so might as well do them in the version they might be released in instead of writing 2 different versions
• Better error handling/typing
  • Before we were assuming cohttp doesn't explode. But it does sometimes. So let's have http helpers return a result on whether or not the network request actually happened, and within that another result on what the response from the server was, if it was ok or not

Semgrep Pro PR

https://github.com/semgrep/semgrep-proprietary/pull/1488

Test plan

Tests pass

[github-actions]

PR checklist:

• Purpose of the code is evident to future readers
• Tests included or PR comment includes a reproducible test plan
• Documentation is up-to-date
• A changelog entry was added to changelog.d for any user-facing change
• Change has no security implications (otherwise, ping security team)

If you're unsure about any of this, please see:

• Contribution guidelines!
• One of the more specific guides located here

[ajbt200128]

It looks like the new version of cohttp relies on opam 2.1, which windows doesn't have :/ so not sure if the cohttp bump part of this PR is possible. I may just bump cohttp back down a version

[r2c-argo]

semgrep-compare-github-qsndg results

Ran benchmark on 38 repositories

The number of findings differs for 3 repos

Whole benchmark is 1.3% faster (a bit of noise is expected)

Relative speed improvement is 1.01 on average

Relative memory improvement is 0.99 on average

[kopecs]

Overall the API updates look good. Some various questions/other minor improvements suggested.

A couple spots I would want to see more docs before merging but other stuff is mostly informative/nits.

[r2c-argo]

semgrep-compare-github-e40oh results

Ran benchmark on 38 repositories

The number of findings differs for 4 repos

Whole benchmark is 4.0% faster (a bit of noise is expected)

Relative speed improvement is 1.01 on average

• 3% of scans are significantly faster
• 3% of scans are significantly slower

Relative memory improvement is 0.99 on average

[r2c-argo]

semgrep-compare-github-p9179 results

Ran benchmark on 38 repositories

The number of files checked differs for 1 repos

The number of findings differs for 3 repos

Whole benchmark is 0.3% slower (a bit of noise is expected)

Relative speed improvement is 1.01 on average

Relative memory improvement is 0.99 on average

[r2c-argo]

semgrep-compare-github-vxc5f results

Ran benchmark on 38 repositories

The number of findings differs for 2 repos

Whole benchmark is 2.9% slower (a bit of noise is expected)

Relative speed improvement is 0.98 on average

• 3% of scans are significantly slower

Relative memory improvement is 0.99 on average

[r2c-argo]

semgrep-compare-github-4y5o0 results

Ran benchmark on 38 repositories

The number of findings differs for 3 repos

Whole benchmark is 1.3% faster (a bit of noise is expected)

Relative speed improvement is 1.01 on average

• 3% of scans are significantly faster

Relative memory improvement is 0.99 on average

[r2c-argo]

semgrep-compare-github-7zrkd results

Ran benchmark on 38 repositories

The number of findings differs for 2 repos

Whole benchmark is 0.6% slower (a bit of noise is expected)

Relative speed improvement is 0.98 on average

• 3% of scans are significantly slower

Relative memory improvement is 0.99 on average

[aryx]

Glad we removed a functor! I hate functors!

[aryx]

the error management seems a bit heavier now ... it's not possible to factorize
a bit so we just handle Ok body = Ok body, and then whatever else as a general error?