chore(instrumentation-pg): have @opentelemetry/semantic-conventions package dependency with caret (^) version to sync with other instrumentation packages

[serkan-ozal]

… so latest minor versions will be able to used

Which problem is this PR solving?

Currently, @opentelemetry/instrumentation-pg package has @opentelemetry/semantic-conventions dependency with direct version (1.27.0), but other instrumentation packages have that with caret version (^1.27.0) of it. And this leads to having multiple copies of the @opentelemetry/semantic-conventions package with different versions in the AWS Lambda Node.js layer.

Short description of the changes

Change @opentelemetry/semantic-conventions package version to ^1.27.0 to be sync with the other instrumentation packages.

[pichlermarc]

this is pinned on purpose since the incubating entrypoint may include breaking changes in minor versions.

See https://github.com/open-telemetry/opentelemetry-js/tree/main/semantic-conventions#unstable-semconv for recommendations on how to handle this case.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 90.80%. Comparing base (4cf7e6f) to head (c71d3ba).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2664      +/-   ##
==========================================
+ Coverage   90.79%   90.80%   +0.01%     
==========================================
  Files         169      170       +1     
  Lines        8061     8070       +9     
  Branches     1646     1646              
==========================================
+ Hits         7319     7328       +9     
  Misses        742      742              

Files with missing lines	Coverage Δ
...elemetry-instrumentation-pg/src/instrumentation.ts	90.21% <ø> (ø)
...de/opentelemetry-instrumentation-pg/src/semconv.ts	100.00% <100.00%> (ø)
...node/opentelemetry-instrumentation-pg/src/utils.ts	97.65% <ø> (ø)

[pichlermarc]

thanks 🙂
cc @maryliag (component owner)

[trentm]

tl;dr: I think we should be using the local copy (src/semconv.ts) in tests as well.

[The following is copied from a Slack discussion to capture this visibly somewhere.]

If one uses the incubating entry point in tests... then the tests will be somewhat of a check that you properly copied the definuitions to your src/semconv.ts.

I'm changing my mind on this one.

Say one is using .../incubating in tests, then the version of the semantic-conventions dep is updated that drops one of the used unstable semconv definitions (say, ATTR_DB_CLIENT_CONNECTION_STATE). The runtime code will be fine, but the tests will fail. We don't want this test failure to be a reason to force the particular instrumentation to change for the new version of Semantic Conventions right then. (Eventually the instrumentations should migrate to newer Semantic Conventions, but it doesn't have to be right when we update the @opentelemetry/semantic-conventions dep for the whole opentelemetry-js-contrib.git repo.)

IOW, I think we should be using the src/semconv.ts in tests as well.

(Ensuring that the copied defs in src/semconv.ts match what is in the actual .../incubating endpoint should, ideally, be the job of a generator tool or a linting script of some sort.)

[trentm]

There was some discussion on Slack to refine this PR. It mimics #2599 which was doing the same thing. Credit to @nwalters512 as well for the changes.