feat: winston instrumentation

[seemk]

Which problem is this PR solving?

Similarly to bunyan (#419), adds trace context to log metadata fields.

Short description of the changes

Injects trace_id, span_id, trace_flags to winston log metadata.

Note that testing with older winston versions (1.1.2 and 2.4.4) was done manually. To help with testing in the future, see open-telemetry/opentelemetry-js#2110

[codecov]

Codecov Report

Merging #425 (8594d81) into main (692932e) will not change coverage.
The diff coverage is n/a.

❗ Current head 8594d81 differs from pull request most recent head 63ea62e. Consider uploading reports for the commit 63ea62e to get more accurate results

@@           Coverage Diff           @@
##             main     #425   +/-   ##
=======================================
  Coverage   95.83%   95.83%           
=======================================
  Files          15       15           
  Lines         624      624           
  Branches       94       94           
=======================================
  Hits          598      598           
  Misses         26       26           

[obecny]

This is really nice piece of work, thx for contribution

Would you be able to include the screenshots to show how the spans look like, can be in this or separate PR ?

[seemk]

This is really nice piece of work, thx for contribution

Would you be able to include the screenshots to show how the spans look like, can be in this or separate PR ?

Perhaps I understood the question wrong, but the instrumentation itself does not generate spans, only injects the active span context to log records. E.g. an example:

console.log(getSpanContext(context.active()));
logger.info('foobar');
logger.info('foobar', { something: 42 });

Without instrumentation:

With instrumentation: