Add pytest-benchmark and port the benchmarks from GraphQL.js

[ktosiek]

This is a start for working on #3 - currently only one benchmark is ported from GraphQL.js, but when this code is in an acceptable shape I'll try porting the rest.

Benchmarks can be run with pytest --benchmark-enable -k benchmark.
By default benchmarks work like simple tests (they only run once), so they shouldn't slow things down too much.

[Cito]

Thanks a lot for this contribution. I always wanted to add the benchmarks - being pushed by others helps a lot. Overall, this looks very good. See below for a few minor suggestions for changes.

[Cito]

We could also use benchmark-skip to completely skip them, but I agree this is better. They are probably a bit redundant, but not too slow if they run only as test.

[ktosiek]

I've ported the rest of the benchmarks.

Impact on the test suite, as measured with time: (Warning: not current, see comment below)

Command	Python 3.6	Python 3.7
pytest	0m11,881s	0m7,113s
pytest --benchmark-skip	0m7,001s	0m4,311s
pytest --benchmark-enable	0m28,314s	0m17,205s

So even running the benchmarks as tests adds 3-4s to the whole test suite. Main offenders are:

1.13s call     tests/benchmarks/test_introspection_from_schema.py::test_executing_introspection_query
0.67s call     tests/benchmarks/test_validate_gql.py::test_validating_introspection_query
0.64s call     tests/benchmarks/test_validate_sdl.py::test_validating_big_schema_sdl

[ktosiek]

Update: there was an optimisation in GraphQL.js test setup that I've missed when porting, here are the current times:

Command	Python 3.6	Python 3.7
pytest	0m10,336s	0m6,125s
pytest --benchmark-skip	0m6,969s	0m4,252s
pytest --benchmark-enable	0m26,654s	0m16,411s

So even running the benchmarks as tests adds 2-3s to the whole test suite. Main offenders are:

0.69s call     tests/benchmarks/test_introspection_from_schema.py::test_executing_introspection_query
0.65s call     tests/benchmarks/test_validate_sdl.py::test_validating_big_schema_sdl

[Cito]

Great job, thanks. See two small issues below. And maybe the test function names should match with the names used in GraphQL.js (specified as const name =....

[ktosiek]

I've renamed the test functions.
I see there's a conflict in poetry.lock - should I handle that? Are there any poetry-specific tools for handling merges?

[Cito]

Thanks. Don't worry, I can resolve that conflict. Will be very busy the next 2 days and look at this later.