Remove surrounding backticks before parsing output as JSON

[irgolic]

gpt-3.5-turbo consistently returns the output surrounded by backticks for me:

E.g.,

```json
{
    "action": "new_file",
    "new_file": {
        "filepath": "tic_tac_toe.py",
        "description": "Python file containing the implementation of the Tic Tac Toe game and AI"
    }
}
\```

(I added the \ on the closing backticks to avoid closing the codeblock)

This PR removes the backticks before JSON parsing, should they appear.

[ShreyaR]

Looks good!

I can approve once tests are passing.

Btw, have you tried out the prompting technique where you most of the prompt in the system messages, and only pass in the low level task details in the user message? Curious if you're still getting backticks with that approach.

[codecov]

Codecov Report

Patch coverage: 60.00% and project coverage change: -0.08 ⚠️

Comparison is base (5210d41) 77.54% compared to head (74de373) 77.46%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #118      +/-   ##
==========================================
- Coverage   77.54%   77.46%   -0.08%     
==========================================
  Files          44       44              
  Lines        2329     2339      +10     
==========================================
+ Hits         1806     1812       +6     
- Misses        523      527       +4     

Flag	Coverage Δ
unittests	77.46% <60.00%> (-0.08%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
guardrails/run.py	90.90% <60.00%> (-3.48%)	⬇️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[irgolic]

Fixed tests, there seems to be some weird behavior that's different on 3.8 and 3.9 in the SQL test I fixed.

Btw, have you tried out the prompting technique where you most of the prompt in the system messages, and only pass in the low level task details in the user message? Curious if you're still getting backticks with that approach.

I've not! Is there an easy way to transition to this system? :D