inconsistency between getsamples() and i_img_bits_t names

[tonycoz]

getsamples()/setsamples() accepts a type value of "float", which was reasonable at the time, but the new data() method will distinguish between returning "float" data (as packed C floats) and "double" data (as packed C doubles).

This inconsistency at the very least complicates testing, but it also makes learning the interfaces harder.

While getsamples() etc won't support packed C floats soon, it may in the future, so keeping them consistent would be nice.