[PROF-9035] Stop worker on clock failure

[AlexJF]

What does this PR do?
When detecting an interaction that involves a clock time of 0 (returned on error), put the worker into a failure state and ultimately lead to it stopping.

Motivation:
We can't trust things if we have non working clocks and better to shut things down than to continue consuming resources.

Additional Notes:

How to test the change?

For Datadog employees:

• If this PR touches code that signs or publishes builds or packages, or handles
  credentials of any kind, I've requested a review from @DataDog/security-design-and-guidance.
• This PR doesn't touch any of that.

Unsure? Have a question? Request a review!

[ivoanjo]

👍 LGTM

[codecov-commenter]

Codecov Report

Attention: 1 lines in your changes are missing coverage. Please review.

Comparison is base (8212e9f) 98.23% compared to head (4a03871) 98.23%.

Files	Patch %	Lines
...filing/collectors/discrete_dynamic_sampler_spec.rb	95.83%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #3439      +/-   ##
==========================================
- Coverage   98.23%   98.23%   -0.01%     
==========================================
  Files        1277     1277              
  Lines       75141    75175      +34     
  Branches     3544     3547       +3     
==========================================
+ Hits        73817    73849      +32     
- Misses       1324     1326       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ivoanjo]

Left a few more notes.

I must say, after going through this PR it's definitely... a very big change to enable the discrete sampler to say "hey, the clock is broken".

I kinda wonder if instead we should move the responsibility of checking the clock back to the CpuAndWallTimeWorker. It already needs to check for errors when calling the discrete sampler so it's not like we're "hiding" a lot of complexity away from it.

On the other hand I kinda realize my suggestion may be ~slightly cruel after you spent all this time on this PR (and I did a bit more on top as I was reviewing it -- the thought only kinda occurred to me at the end).

Also, the current mechanism is definitely more flexible in terms of allowing many more error states to be introduced, whereas my suggestion only really solves the issue of the clock error.

I leave the decision up to you ;)

[ivoanjo]

👍 LGTM