Feat/nodejs/add auto instrumentations

[wpessers]

This PR fixes #1723, fixes #1600, fixes #1611
Added all the instrumentations that exist in the otel auto-instrumentations-node metapackage. Also removed most of the defaults. Right now only defaults (besides aws-lambda and aws sdk instrumentation) left are:

dns
http
net

I can see 2 possible changes to this:

1. Remove dns and net from the defaults. This was already discussed before in the community, and I personally also question them being enabled by default. I think opt-in for these may make more sense.
2. Add undici to defaults. I'm personally not 100% sure about this one. This would fix users not seeing traces when using nodejs native fetch. But then again, not everyone uses native fetch.

As of now the PR is still a work in progress. I'm still working on a proposal to refactor the logic for the dynamic imports. As the wrapper code has gotten a bit bloated with all newly added instrumentations: wrapper.ts

Also, as this was discussed in the otel faas working group. The changes do indeed not noticably impact package size and cold starts. After adding the extra instrumentation packages, the package size increased by ~40KB. I ran some tests and noticed no difference in cold start times.

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: wpessers (fa350e2, c05a96c, 147d871, 865428a, 3091298, 763ef0c)

[wpessers]

Separate comment for readability. After discussion with maintainers I've reverted the change where we alter the default instrumentations. This PR's purpose is only to support more out-of-the-box instrumentation and therefore we don't want it to be breaking. Removing defaults is, however, breaking so that will be a separate PR.

[wpessers]

As requested I also ran some quick tests to see impact on the actual OTEL wrapper initialization times.

• In my testing, one lambda uses the OTEL_NODE_ENABLED_INSTRUMENTATIONS env var to only load the http instrumentation. I will refer to this lambda as the light lambda.
• The other one does not set the env var and thus loads all instrumentations, I will refer to this lambda as the heavy lambda.

These are the results after running 40 times (data in ms):

Statistic	light_old	light_new	heavy_old	heavy_new
Min	194	196	230	219
Max	260	265	358	341
Median	211	228	253	270.5
Mean	214.475	228.525	259.7	271

Here's the raw data: otel_wrapper_init_times_ms.csv

Sample size n=40 may not be a lot. But these results do look assuring that the changes have minimal to no impact on wrapper initialization time. If further testing is required, I can of course alter the parameters and/or sample size to get more data.

[wpessers]

Fixed @serkan-ozal's comment about cucumber instrumentation. Also removed instrumentation for generic-pool and lru-memoizer as these don't make sense to use in lambda functions.

[pragmaticivan]

LGTM!

[pragmaticivan]

Btw, it seems the failures in check-links are legit, so we probably need to fix these links later in a new PR. Not related to this PR though.