feat: add dbtext decompression benchmark

[NickCondron]

1. Is decompression throughput typically defined as bytes/s of input or output?
2. Decompression's measured speed is fast, but the iterations are slow due to large drop. Criterion complains that it doesn't get 100 samples in 5s. I could increase the sample time or reduce the minimum samples. If there was a analogous decompress_into function the memory could be reused and this wouldn't be an issue.

[CLAassistant]

All committers have signed the CLA.

[a10y]

@NickCondron thanks for contributing! A few points to address, see my comments.

As I mentioned, the decompression throughput is canonically measured in bytes-out per unit time.

These are the bench results I get on my M2 Max laptop:

[a10y]

And here are the numbers from a 10 year old Intel laptop I have lying around, which echoes the 1-3GiB/sec that the FSST paper authors seemed to get

[NickCondron]

I appreciate the detailed comments. I pushed your recommended changes. I never figured out how to get the only the decompression benchmarks to run, so I didn't run into that issue initially haha.

[a10y]

Thanks!