*: update parquet-go to include goroutine leak fix

[asubiotto]

This should alleviate our CI flakes where too many goroutines are alive at one
time.

There is also a fix pulled in for buffering bloom filter reads.

[asubiotto]

Interesting failure mode on first CI run. Looks like an unrelated flake:

=== RUN   Test_Table_Concurrency/8192
race: limit on 8128 simultaneously alive goroutines is exceeded, dying
FAIL	github.com/polarsignals/frostdb	12.551s

The test only seems to spawn 8 goroutines. Trying to reproduce the failure locally.

[asubiotto]

Looks like the cause if a bunch of leaked goroutines in asyncPage.init:

         Goroutine 105824 in state select, with github.com/segmentio/parquet-go.readPages on top of the stack:
        goroutine 105824 [select]:
        github.com/segmentio/parquet-go.readPages({0x105d992a0, 0xc06f880198}, 0xc06f23e180, 0xc06daf9ab0, 0xc06f23e0c0)
                /Users/asubiotto/go/pkg/mod/github.com/segmentio/parquet-go@v0.0.0-20220830163417-b03c0471ebb0/page.go:222 +0x168
        created by github.com/segmentio/parquet-go.(*asyncPages).init
                /Users/asubiotto/go/pkg/mod/github.com/segmentio/parquet-go@v0.0.0-20220830163417-b03c0471ebb0/page.go:161 +0x270
        ]

Looks like we're spawning 100k goroutines in a single test run. I think the likeliest thing is that this is a latent bug in our page lifecycle management that has only surfaced due to segmentio/parquet-go#297.

[asubiotto]

The last commit fixes the asyncPage.init leak but CI is still failing. We are still leaking some goroutines (see #180) but this is an issue on main as well. I'm guessing that we're still spawning a ton of goroutines simultaneously when writing rows. I'm wondering if this is expected.

[asubiotto]

Apologies for the excess notifications. I'm going to try peppering the code with some missing Close calls to see if we can at least alleviate the number of simultaneously alive goroutines.

[asubiotto]

I found a potential goroutine leak in parquet (fixed in segmentio/parquet-go#337) which I think fixes our problem (at least makes the leak detector not scream at me when running without finalizers). I'll merge the upstream fix, update the version on this PR and then retry.

[asubiotto]

RFAL, I pulled in the upstream goroutine leak fix as well as the bloom filter read buffering @thorfour