refactor(s2n-events): Reduce generated diffs when changing metric definitions

[Mark-Simulacrum]

Release Summary:

• feat(s2n-quic-dc): Instrument connect_tls with latency/error event tracking

Resolved issues:

n/a

Description of changes:

Previously the raw integers were embedded in multiple places, and inserting into the middle of the list caused all integers to change. This meant a huge diff for a tiny change, and also increased the risk of merge conflicts. This updates the generator to have a list of constants each with value (previous + 1), which means that any change should only touch ~1 adjacent constant to any newly added metric at most.

This shouldn't have any change in runtime behavior or performance since all offsets/indices remain computed at compile-time.

Call-outs:

n/a

Testing:

There's some unit/integration tests and I've poked through a couple of example events and things look reasonable.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[Mark-Simulacrum]

I'll rebase this once the generator is approved since it needs rebasing after any event-touching PR.

[boquan-fang]

Is there a diff that you can show for this PR? It might be good to have a example of it to show how the generated diff is reduced.

[Mark-Simulacrum]

I pushed a second commit adding an event around s2n-quic-dc TLS stream connects for an example diff. It would've produced a much larger diff (standard ~3k lines changes, see e.g. 4ca17fa for previous example of that kind of change).