Closes #2472, #2800: Update IndexingMsg to use & instead of mod for setting bigint pdarrays

[jaketrookman]

This PR (Closes #2472 and closes #2800) updates the function calls of mod in IndexingMsg to use %, as they are more efficient.
The bool case still needs to be updated as there is not a way to directly cast a bigint to a bool.

Leaving as a draft until I know if this last case can be completed without this casting feature.

[stress-tess]

@jaketrookman this isn't really what the issue was getting at

For this instead of using % or .mod, i was think we use &. This is what we do in other sections of code and it's more efficient. Basically instead of val % 2**max_bits we do val & ((2**max_bits)-1) so it functions as a bit mask instead

You can see something similar to this done in operatorMsg

var max_size = 1:bigint;
var has_max_bits = max_bits != -1;
if has_max_bits {
  max_size <<= max_bits;
  max_size -= 1;
}
...
if has_max_bits {
  li &= local_max_size;
}

Also just looking at this code statically, unless im misreading something what's already there seems to be incorrect. Because it's doing val % max_bits not val % 2**max_bits. Let me look into this a bit

[stress-tess]

yup the existing code (which i wrote 🙃 ) is incorrect when max_bits is set... see reproducer below. This is broken on master and is captured in issue #2800. We def want this corrected and tested before the next release

In [3]: bi = ak.arange(2**200 - 1, 2**200+4)

In [4]: bi
Out[4]: array([1606938044258990275541962092341162602522202993782792835301375 1606938044258990275541962092341162602522202993782792835301376 1606938044258990275541962092341162602522202993782792835301377 1606938044258990275541962092341162602522202993782792835301378 1606938044258990275541962092341162602522202993782792835301379])

In [5]: bi[:] = ak.arange(2**200 - 1, 2**200+4)

In [6]: bi
Out[6]: array([1606938044258990275541962092341162602522202993782792835301375 1606938044258990275541962092341162602522202993782792835301376 1606938044258990275541962092341162602522202993782792835301377 1606938044258990275541962092341162602522202993782792835301378 1606938044258990275541962092341162602522202993782792835301379])

In [7]: bi.max_bits = 201

In [8]: bi
Out[8]: array([1606938044258990275541962092341162602522202993782792835301375 1606938044258990275541962092341162602522202993782792835301376 1606938044258990275541962092341162602522202993782792835301377 1606938044258990275541962092341162602522202993782792835301378 1606938044258990275541962092341162602522202993782792835301379])

In [9]: bi[:] = ak.arange(2**200 - 1, 2**200+4)

In [10]: bi
Out[10]: array([3 4 5 6 7])

[stress-tess]

one nit pick on the tests but the logic looks great!

[stress-tess]

Looks good!