Implement snp.broadcast_shapes

[Michael-T-McCann]

There are times when we would like to know the result of broadcasting a BlockArray against another BlockArray. Note that the docs should already define how this broadcasting will work.

Currently, we do this with

scico/scico/linop/_linop.py

Line 196 in 5ae48c4

output_shape = (snp.empty(input_shape) * diagonal).shape

Ideally, we could do this by implementing snp.broadcast_shapes so that it works correctly with nested shapes. What makes this hard is that scico.numpy (1) has some complex metaprogramming and (2) circular import problems that stop you from accessing utils.is_nested.

For more context, see #144.