feat: Implement native st_pointz, st_pointm, st_pointzm

[petern48]

• Implements native functions st_pointz, st_pointm, st_pointzm
• Cleans up and makes that file more readable (previously a lot of magic numbers)
• Add support for ArrayArrayArray and ArrayArrayArrayArray benchmark args

[paleolimbot]

We have some new helpers that I think will make the implementation a bit more concise that are worth a try...the test code looks great!

[paleolimbot]

The inside of this loop is a very performance-sensitive section and I think there are a few things we can do to help:

for ... {
  let values = (0..num_dimensions).iter().map(|j| coord_f64_arrays[j].value(i).collect::<Vec<_>>();
  let values_non_null = values.iter().take(num_dimensions).flatten().collect::<Vec<_>>();
  if values_non_null.len() == num_dimensions {
    write_wkb_point_header(&mut builder, dimensions)?;
    // I forget exactly how to write the f64s as little endian doubles to the builder but it's
    // the same as in write_wkb_xxxx()
   builder.append_value([]);
  } else {
    builder.append_null();
 }
}

This was the first ever function I wrote in sedonadb and we now have write_wkb_point_xx(). The builder implements Write, so you can write directly into it (pass it directly to write_wkb_xxx() functions).

Make sure you run cd rust/sedona-functions + cargo bench -- st_point to make sure your implementation doesn't slow down the XY case (the most common). It is probably worth separating the implementation completely for Z, M, and ZM (where performance is not as critical).

[petern48]

Yeah, you were right. It was a notable perf regression (at least 100%), so I reverted st_point.rs back to the original version and moved everything else over into a new `st_pointzm.rs

[paleolimbot]

Thank you!