Make brotli configurable by environment variable

[ddelange]

This PR allows configuring compression level and compression mode via environment variables.

• Django's default gzip compression level is 6 out of 9.
• Brotli's default compression level is 11 out of 11.
• This PR sets django-brotli's default to 4 out of 11.

For a response with nearly 2MB uncompressed json payload containing many lists and dicts of english words (broti's strength), integers, and floats, a brotly quality setting of 5 or lower is needed avoid being slower than the gzip middleware (and still achieve 20% better compression ratio).

Setting the default quality to 4 makes brotli twice as fast as gzip@6, and still yields 20% better compression.

Benchmark:

>>> payload = open('./payload.json', 'rb').read()
>>> len(payload)
1759385
>>> import gzip, brotli
>>> len(gzip.compress(payload, 6))
516307
>>> len(brotli.compress(payload, quality=11))
328070
>>> len(brotli.compress(payload, quality=10))
336622
>>> len(brotli.compress(payload, quality=9))
361724
>>> len(brotli.compress(payload, quality=8))
366235
>>> len(brotli.compress(payload, quality=7))
371714
>>> len(brotli.compress(payload, quality=6))
381026
>>> len(brotli.compress(payload, quality=5))
394309
>>> len(brotli.compress(payload, quality=4))
419936
>>> len(brotli.compress(payload, quality=3))
483480
>>> len(brotli.compress(payload, quality=2))
499352
>>> %timeit gzip.compress(payload, 6)
77 ms ± 1.07 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=11)
4.16 s ± 137 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
>>> %timeit brotli.compress(payload, quality=10)
1.74 s ± 74.1 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
>>> %timeit brotli.compress(payload, quality=9)
178 ms ± 15.5 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=8)
134 ms ± 9.4 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=7)
111 ms ± 7.22 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=6)
80.9 ms ± 2.43 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=5)
74.9 ms ± 1.23 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=4)
38.2 ms ± 431 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=3)
24.6 ms ± 309 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)
>>> %timeit brotli.compress(payload, quality=2)
21.3 ms ± 773 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)

[ddelange]

Hi @illagrenan 👋 could you have a look if you have a minute?

[illagrenan]

Thanks for your pull request, it looks good :-)