32-bit integer overflow of so3g.block_minmax

[ykyohei]

This leads to Segmentation fault (core dumped).

import so3g
import numpy as np
a = np.zeros((4050, 603260), dtype='float32')
so3g.block_minmax(a, a, 800, 2, 0)

This is used for slow_jumps of sotodlib
https://github.com/simonsobs/sotodlib/blob/a6176f052d392019b50dc85b994ba47cf5098058/sotodlib/tod_ops/jumps.py#L601
When we analyze single wafer, the number of dets x samps is small enough and we do not have segmentation fault, but we have issue when we analyze multiple wafers at the same time.

I seem to have same issue for so3g.subtract_jump_heights etc
https://github.com/simonsobs/sotodlib/blob/a6176f052d392019b50dc85b994ba47cf5098058/sotodlib/tod_ops/jumps.py#L211

[ykyohei]

This part of codes seems to be written by @skhrg

[skhrg]

Sorry just saw this. In your example there you have ndet = 4050 and nsamps = 603260 so ioff overflows.

I'll put in a PR to use 64-bit ints and the indices which should fix this.

[ykyohei]

I reopen this until we really fix it.