align_corners is handled incorrectly in RandZoomd (and possibly bugs in other transforms)

Example:
RandZoomd(keys=['image'], prob=0.5, min_zoom=0.9, max_zoom=1.1, mode=['trilinear'], align_corners=[False])

Will call Zoom transforms with mode = trilinear, but align_corners=None. Which will also show a lot of pytorch Warnings indicating that align_corners is not set.

Root cause:

when calling Zoom transform, for some reason we use parameters as

mode=look_up_option(mode or self.mode, InterpolateMode),
align_corners=align_corners or self.align_corners,

The use of 'or' is very ambiguous and leads to a bug in this case.

A or B -> returns A (if both A and B are strings or enums)
A or B -> returns B (if either A==None or B==None)

In our case
mode or self.mode -> returns mode
align_corners or self.align_corners -> returns self.align_corners (because align_corners==None)

as a result we get mode='trilinear', align_corners=None (the original option is lost)

We use this 'or' throughout the code for align_corners, and other parameters, which potentially created bugs if one of the options is None...