I think it would be good to explicitly explain why we have all these catagories of types. I'd replace the sentence above with this text:

NumPy provides both bit sized type names and names based on the names of C types. Since the definition of C types are platform dependent, this means the explicitly bit sized should be preferred to avoid platform-dependent behavior in programs using NumPy.

To ease integration with C code, where it is more natural to refer to platform-dependent C types, NumPy also provides type aliases that correspond to the C types for the platform.

It would also be good to explain why some types don't have a canonical name or don't have a C-like name.

Then I think for bool would be that bool does not exists and 0 and 1 ints are preferred (or <stdbool.h> is required).
For intp/uintp the same name could be considered as "C-like".
For longlong, because it only a C concept ported to Python?

I think it's ok to use the stdbool types in the table, we're already using other c99 types.

I think I just wanted you to comment that some of the type names have a trailing underscore to avoid confusion with builtin python type names. No need to go into detail on longlong and intp.

Right, I added a comment about trailing underscore.