Allocate before loading when responding to queries

[Stebalien]

In the query executor, we seem to load first, then allocate when sending the block. Unfortunately, when processing many parallel queries, we can end up loading a lot of blocks then blocking (waiting to be able to "reserve space" for them. we'd likely use less memory if we just sent the block.

Instead, we should allocate first. That is:

1. Before loading anything, allocate 2-4MiB (maximum block size).
2. Then load.
3. Then reduce the allocation to the actual amount of memory needed.

This means we:

1. Won't go over the allocation limit.
2. If we're blocked on allocating, we won't sit there while holding on to buffers.

[hannahhoward]

Yea this makes sense, but I'm not sure how much it will help us. My feeling is the massive spikes we're seeing in Estuary are not off by 1 loads but memory getting held for some reason.

It is still worth doing.

[hannahhoward]

For anyone picking up this ticket: you will need to move use of the allocator around a bit to do this. You'll need to pipe it through as a dependency to the query executor, and possibly run traversal.