Unclear why brotli is disabled by default

Nobody has claimed this yet.

Hi,

I see that the [documentation](https://github.com/lukeed/polka/tree/02cbdb529ddca0a9f3d225e2abb2931924219cc3/packages/compression#optionsbrotli) states:

> Brotli incurs more performance overhead than Gzip, which is why it's not enabled by default.

But looking at most files on https://quixdb.github.io/squash-benchmark/ (e.g. `plrabn12.txt` on `s-desktop`), that doesn’t seem to be the case at what polka states to be their default settings (6 for both, although [Brotli’s documentation](https://www.brotli.org/encode.html#a8d0) says the default is 11). Alternatively, https://github.com/powturbo/TurboBench finds that on their corpus, `brotli -q 4` compresses faster than `gzip -6` and still denser than `gzip -9`. Is it perhaps the case that Brotli was disabled in polka based on its speed at `-q 11` being mistaken for its speed at `-q 6`?

(Disclaimer: I happen to work at Google on the team that originated Brotli before I joined, but I am writing this in my personal capacity. Working with them just happens to have sensitised me to the popular misconception that Brotli is inherently slow.)

Start with the documentation at packages/compression#optionsbrotli and compare its stated defaults and performance rationale with the benchmark links in the issue. Verify whether the explanation accurately reflects Polka’s behavior, then clarify the documentation; done when the reason Brotli is disabled by default is unambiguous.

Written by the indexing model from the issue text.