Implementation of moving quantile instead of moving median, which is a partial case of quantile. For now using the fixed constant in #define

 into quantile

Add quantile and has_quantile parameters to the template

Added all imports
Fixed a bug when q=1. Call move_max for this case (on python layer)
Added a lot of tests for move_median and move_quantile

Fix keyword argument "method"/"interpolation" for different
numpy verisons (keyword was changed after 1.22.0)
Copy over doc string for move_quantile from C layer to Python layer

This reverts commit c2a2ae3.

move_min is significanlty faster than move_quantile with q = 0. So in case of
q=0 apply move_min instead. Same for q=1 and move_max.

Instead of move_quantile substituting move_median completely,
have both move_median and move_quantile implemented
separately.

Remove the wrapper for move_quantile
on python level which checked for
q = 0 or 1. Now it's fully in C.
Also check for q=0.5 as we checked
it's 3-4% faster to call move_median

Add some more tests

Remove redundant import

Mostly get rid of macros

for versions comparison

This eliminate the need for macro in move_median.c
mm_handle will just have an unused membet "quanitle"
for the case of move_median.

move_median and move_quantile now have all the same
functions except for the construction of mm/mq.