Revert resource tracking to once again print to stdout

[jzaia18]

Context:
The new resource tracking feature introduced in #7226 logs resource tracking information to a json file. This can cause issues when the file gets clobbered after multiple runs.

Description of the Change:
This PR reverts the behavior of resource tracking to once again print resource information to stdout instead of writing to a json file. This behavior is consistent with Catalyst.

Benefits:
Avoids potential issues with file clobbering.

Possible Drawbacks:
Dirties stdout when resource tracking is enabled, makes parsing output slightly more involved.

Related GitHub Issues:
#7226
[sc-88213]

[github-actions]

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

• A one-to-two sentence description of the change. You may include a small working example for new features.
• A link back to this PR.
• Your name (or GitHub username) in the contributors section.

[albi3ro]

If we're running into issues with the file getting too much stuff in it, could we just log each run to a different file by allowing the file to be set on the device at initialization? I can't see how writing to stdout would keep things cleaner than a file.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.65%. Comparing base (e21660f) to head (480fd93).
Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #7392   +/-   ##
=======================================
  Coverage   99.65%   99.65%           
=======================================
  Files         529      529           
  Lines       50610    50615    +5     
=======================================
+ Hits        50434    50439    +5     
  Misses        176      176           

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[maliasadi]

🙌

[AntonNI8]

Looks good, pending existing suggestions👍

[josh146]

I haven't been involved in the discussions here, so apologies if this was already discussed, but is there not a way here to support structured and/or configurable output from null.qubit?

E.g., an option to null.qubit for users to provide the print behaviour (stdout, logger, writing to a file, etc).

[jzaia18]

I could do this but it would likely involve adding another parameter to the constructor. Alternatively I could replace track_resources as a bool with something like (None => do not track, file-like => use the provided object, str => create a file with this name). Thoughts? @josh146 @albi3ro @maliasadi

[josh146]

@jzaia18 I'll leave this to others that are involved in this project, but I can imagine as a user common use cases will be wanting to perform computations/visualizations on the data, and it can be non-trivial to do this programmatically if the data is on stdout!

As a python user I would instead want to work with structured data of some sort (dicts, np arrays, etc)

[AntonNI8]

As a python user I would instead want to work with structured data of some sort (dicts, np arrays, etc)

I'm not concerned with what's end-user-facing here, but it's true that the flexibility could be helpful internally.
The default to stdout to avoid JSON files blowing up, and for consistency with Catalyst makes sense.
Adding another parameter to the constructor seems better than trying to encapsulate all of the details in track_resources.

[jzaia18]

I have a working prototype that takes the following approach. track_resources may be one of (bool, None, str, file-like). If False or None, resource tracking is disabled. If True, it writes to stdout. If given a filename as a str, it creates and writes to the passed file. And I don't advertise this in the docstring because I can't easily support this with Catalyst, but if you pass a file-like object, it can write to that too. All other options I was able to test and get working with Catalyst. I think this may be the closest to a solution that satisfies everybody, and it should make my hook in the benchmarks repo more streamlined as well. What do you all think? @maliasadi @AntonNI8 @josh146