Explicitly cast float to mp_int_t

[dunkmann00]

Not sure why this is necessary, but it prevents an off-by-one error in some (rare?) circumstances.

Fixes the bug found in #5725:

>>> import board, neopixel
>>> pixels = neopixel.NeoPixel(board.D5, 8, brightness=1.0, auto_write=False, pixel_order=neopixel.GRBW)
>>> pixels[2] = 0x400000FF
>>> pixels[2]
(0, 0, 255, 0)

This doesn't close #5725, since that it is also about whether or not the behavior of integer values with RGBW neopixels is appropriate.

[jepler]

Thank you for finding this! It is "obvious in retrospect", for at least some definitions of obvious, but I had looked directly at this code just recently and still not seen it for myself.

I'm approving this PR, but with the caveat that we may wish to have the bugfix in the 7.1.x branch. @dhalbert please advise on that point.

Here's a rough expression based on my understanding of C: When using the conditional operator such as cond ? x1 : x2, the type of the whole expression depends on the types of x1 and x2. Here, x1 is some kind of integer and x2 is some kind of float. This means that the expression as a whole is of float type (due to the "usual arithmetic conversions", the same as for code like x1 + x2, making x1 be cast to a float and then the result of the whole conditional operation is cast back to an int. By adding the cast to the x2 expression, both x1 and x2 are integer types, so x1 doesn't undergo any undesirable implicit casting.

[dhalbert]

I will backport this to 7.1.x, no problem.

[dunkmann00]

@jepler Interesting! It took me a while to figure out that the bug was simply coming from the inline conditional so I very much appreciate that explanation as to why. Thanks!