chore: Add initial benchmarks for Arrow IPC Stream writer

[andygrove]

Which issue does this PR close?

Part of #6968

Rationale for this change

Add some initial benchmarks to help with efforts to optimize IPC reading and writing.

This is a copy of existing code that we had in Comet and is not as comprehensive as the suggestions in #6968 but it provides a starting point at least.

arrow_ipc/write_single_batch
                        time:   [20.784 µs 20.947 µs 21.091 µs]

arrow_ipc/write_multiple_batches
                        time:   [196.58 µs 199.16 µs 201.40 µs]

What changes are included in this PR?

Are there any user-facing changes?

[alamb]

Thank you @andygrove

I tried this out locally:

cargo bench --bench stream_writer

I also made a flamegraph with

 sudo flamegraph -- target/release/deps/stream_writer-a9122e21216257a4 --bench

I have a suggestion for improvement (don't buffer) but that could be done as a follow on

[alamb]

I ran a quick flamegraph of this benchmark and a non trivial amount of time was spent resizing this buffer:

I recommend we resize / reuse it so we aren't testing the speed of the allocator

Something like this:

• reuse target buffer in icp writer benchmark andygrove/arrow-rs#76

[alamb]

Interestingly, when I resized the output buffers reasonably in andygrove#76 and reran the flamegraph, a substantial amount of time is still spent reallocating / resizing:

Here is the whole file:

[tustvold]

I believe this is waiting on andygrove#76 so marking as draft