Track unknown errors

[Nishnha]

Dependabot currently treats unknown_errors the same as any other error type.

However, unknown_errors are a bit of a special case because they are mostly caused by a HelperSubprocessFailure when Dependabot shells out to native package managers. The helper subprocess failures are harder to debug because they can be specific to a dependency set or some combination of a package manager version and project file features.

In order to better investigate unknown errors, we should send the error details to the backend service so we can look for trends.

This PR makes a few changes to the way unknown errors are logged:

1. unknown errors can only be generated from a few sources:
   a. helper subprocess failures
   b. file fetcher command errors
   c. update files command errors
2. helper subprocess failure commands are now logged to the Dependabot job log so users can investigate issues themselves (Improve error when native helper gives empty output #6521)
3. octokit 500 errors are treated as a new error type, server_error, since they are well-known
4. unknown errors also increment an updater.unknown_error metric with tags for the package manager

Still needs a feature flag for GHES support and tests for the new endpoint.

EDIT: Check has been added in Dependabot-api to to return immediately if in GHES or Proxima environment

[jurre]

The changes look good to me, but I am slightly worried we could be loggin PII when logging the commands we're running natively, so it might be good to have a look at that. Going to approve as to unblock this once we're confident about that, though 👍