[GA-1525] NCL Entry for Precision Differences

[cyschneck]

NCL uses a different floating type which impacts precision. Would be useful to make a fully entry about the origin of the precision differences to be able to link to in future entries

For example:

asin: 
	python (math):	0.5235987755982989
	python (numpy):	0.5235987755982989
	ncl:		0.5235988

There are some diffs in the definitions depending upon the context.

In NCL the float type is 32 bits and double 64 bits, whereas in Python the float type is generally a 64-bit floating point number.

This is also slightly different from numpy which has its own types (e.g. np.float64, np.float32).

Generally the 64-bit types are the defaults in Python though, which is different from NCL and something that's come up in the past.

(Originally from #40)

[kafitzgerald]

We might also think a bit more about some good defaults and/or guidance for setting tolerances (i.e. in math.isclose() or numpy.isclose()) in our receipts testing.

This could be it's own chunk of work, but I think it aligns here thematically.

This came up most recently in Andy's climatology PR. Conversation here: #80 (comment)

Some relevant docs / reading:

• https://peps.python.org/pep-0485
• https://numpy.org/doc/stable/reference/generated/numpy.isclose.html
• https://docs.python.org/3/library/math.html#math.isclose