Compute time to transfer `n` points.

The best way to stream points in the point budget is to know how many points we can transfer easily per frame. This is fairly trivial to compute:

1. The client asks for 10 points from the server. Immediately before requesting points, set a timestamp. When the response is completed, set another timestamp. Save this value $t_1$ as the baseline.
2. The client asks for 10000 points from the server. Do the same process, saving in $t_2$.
3. The client asks for 1000000 points from the server. Save the result in $t_3$.

After a long conversation with ChatGPT we figured out a solution:

$t_i = n_i * m + o$ where m is the time to transfer 1 point and o is the overhead.

$m = (t_2 + t_3 - 2 * t_1) / (n_2 + n_3)$ and $o = t_1 - n_t * m$

https://github.com/rapidsai/node/issues/444#tasklist-block-a45d0a65-ade6-4768-9417-a63367447c89

No source files, tests, or entry points are named in the issue, so first locate the client/server point-transfer path and its timing or streaming logic. Reproduce requests for 10, 10,000, and 1,000,000 points, then verify that the measured transfer-time model can provide a usable per-frame point budget.