Updated telemetry events

[akoutmos]

This PR refactors telemetry events so that they are in their own modules and used in the applicable spots. This cleans up the the consumer and publisher modules and pulls documentation into the modules themselves. This design is largely influenced by Finch (https://github.com/keathley/finch/blob/master/lib/finch/telemetry.ex).

I also updated the fields in the measurement maps to align with the comments in beam-telemetry/telemetry#57.

Description

Telemetry events have been updated to be inline with Telemetry guidelines.

Checklist

• I have added unit tests to cover my changes.
• I have improved the code quality of this repo. (refactoring, or reduced number of static analyser issues)
• I have updated the documentation accordingly

[coveralls]

Coverage decreased (-0.4%) to 89.967% when pulling 4248587 on updated-telemetry-events into 7d88f8b on master.

[akoutmos]

This one should be good to review now. All changes that I could gather from the various Elixir community projects are in.

[vorce]

Nice. I think we might have to mention the change from time to system_time as key for the start events somewhere (maybe release notes is enough @mkorszun ?).

[mkorszun]

@vorce yes, will take care of the communication. There will be followup PR to update all docs and maybe document migration steps for 2->3

[mkorszun]

I am not convinced with removing this documentation. We should still treat these modules as internal ones. Users should not interact with them directly meaning that they might not come across their documentation.

[akoutmos]

Good point! I added it back in the README and have the links going to the modules. That work?

[mkorszun]

@akoutmos yes, sounds good 👍

[mkorszun]

@akoutmos could you elaborate a little about changes in telemetry events? I went quickly through beam-telemetry/telemetry#57 but I do not find it clear.

[akoutmos]

@akoutmos could you elaborate a little about changes in telemetry events? I went quickly through beam-telemetry/telemetry#57 but I do not find it clear.

From what I gathered from the various issues and PRs to the other repos, the primary changes were the switch from time -> system_time in the start event. The standardization of the exception event when an exception occurs and a stacktrace is possibly available. Also the fact that the stop event can have an optional error field to denote that something went wrong...but not necessarily an exception with a stacktrace.

[akoutmos]

This one should be good to merge now I imagine ? The CI error is because 1 of the jobs failed to publish to coveralls and the link check will fail until the docs make their way to master.

[mkorszun]

@akoutmos great, will merge it soon. Great job 👍