Use orjson instead of json/ujson

[masipcat]

I'm still testing and doing stress benchmarks to ensure there is a performance gain

[codecov-io]

Codecov Report

Merging #940 into master will decrease coverage by 0.1%. The diff coverage is 73.9%.

@@           Coverage Diff            @@
##           master    #940     +/-   ##
========================================
- Coverage    94.5%   94.5%   -0.0%     
========================================
  Files         333     333             
  Lines       29550   29549      -1     
========================================
- Hits        27909   27903      -6     
- Misses       1641    1646      +5     

Impacted Files	Coverage Δ
guillotina/renderers.py	82.9% <63.0%> (-3.9%)	:arrow_down:
guillotina/request.py	86.0% <87.5%> (-0.6%)	:arrow_down:
guillotina/contrib/redis/dm.py	100.0% <100.0%> (ø)
guillotina/exceptions.py	87.9% <100.0%> (ø)
guillotina/tests/test_ws.py	100.0% <100.0%> (ø)

[masipcat]

Benchmarks

Using molotov:

• 60% create item. JSON body (3300 bytes)
• 40% endpoint calls json.loads() and dumps() of a big JSON (53755 bytes)

	json (reqps)	orjson (reqps)	diff
#1	5064	5661
#2	5423	5587
#3	5107	5401
Mean	5198	5549.7	+6.77%

Using molotov:

• 40% create item. JSON body (3300 bytes)
• 60% endpoint calls json.loads() and dumps() of a big JSON (53755 bytes)

	json (reqps)	orjson (reqps)	diff
#1	5095	5786
#2	5148	5986
#3	4816	5925
Mean	5019.7	5899	+17.52%

So, we have a small improvement (+6%) for small json data and a relatively big improvement for big json data (+17%).

[vangheem]

Sweet! It looks fine to me. Do we want b/w compat GuillotinaJSONEncoder still?

[masipcat]

Sweet! It looks fine to me. Do we want b/w compat GuillotinaJSONEncoder still?

I think we can add it. Is just 3 lines...

[masipcat]

Ready for review!