Postpone request decoding

[lundberg]

This PR aims to enhance the performance and speed of RESPX by postponing instantiation of httpx.Request to places actually needed and used.

TODO:

• Postpone Request instantiation for callback matcher
• Postpone Request instantiation for content callback
• Postpone Request and Response instantiation when recording call stats Fixed in Lazy request and response decode in CallList  #92

Fixes #90

[codecov-io]

Codecov Report

Merging #91 into master will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master       #91   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           11        11           
  Lines         1210      1214    +4     
  Branches        80        80           
=========================================
+ Hits          1210      1214    +4     

Impacted Files	Coverage Δ
respx/models.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update aaac1c3...9a2f85f. Read the comment docs.

[lundberg]

About postponing the instantiation of Request and Response for the call stats, I think it could be moved from the transport.record method to the Call class. Agree @SlavaSkvortsov ?

This would enforce us to record the raw httpcore request and response tuples in the stats magic mock, and then handle decode_request and decode_response when unpacking a call tuple or accessing it.

Code that should trigger and return decoded requests and responses:

request, response = respx.calls[-1]  # also iter calls should return decoded 

last_call = respx.calls.last
last_call.response
last_call.request

Also worth mentioning, the decoded response holds a reference to the decoded request, which votes for local call knowledge of already instantiated request. Meaning, if one accesses response, the request needs to be decoded, and if request is accessed and asserted after that, the request should not need to be instantiated again. Maybe always decode the response on call access and have the request "property" as a shortcut to response.request?

[SlavaSkvortsov]

I'll take a look on it today