Exp: Testing superjson performance

[RobinTail]

This thing seems to be involved into tRPC protocol according to #2180

Claimed by superjson documentation:

• Negligible runtime footprint

Fact checking:

   ✓ Experiment on serialization (2) 1350ms
     name                                    hz     min     max    mean     p75     p99    p995    p999     rme  samples
   · JSON.stringify()                328,889.59  0.0028  0.1905  0.0030  0.0030  0.0047  0.0049  0.0267  ±0.25%   164445   fastest
   · JSON.stringify() with replacer  236,611.42  0.0039  0.3153  0.0042  0.0041  0.0055  0.0064  0.0308  ±0.32%   118306
   · superjson.serialize()            62,360.65  0.0146  0.3626  0.0160  0.0156  0.0326  0.0451  0.2140  ±0.66%    31181   slowest

 BENCH  Summary

  JSON.stringify() - tests/bench/experiment.bench.ts > Experiment on serialization
    1.39x faster than JSON.stringify() with replacer
    5.27x faster than superjson.serialize()

Related issue flightcontrolhq/superjson#68

In general, I admire the idea, but it seems to have a massive performance flaw that makes its usage in IO operation questionable. However, it's more powerful and fixes several issues of JSON methods, like circular references.

[coveralls-official]

coverage: 100.0%. remained the same
when pulling efe99a6 on try-superjson
into 9a2c66e on master.