Include Nyquist bin, halve it for even lengths (Morse)

[OverLordGoldDragon]

Addresses #13.

Note that for odd lengths, not halving last bin (non-Nyquist) is proper and preserves analyticity, but loses fast decay. This decay seems irrelevant for CWT, but not if otherwise working directly with time-domain wavelet. This is conveniently worked around if CWT pads up to next-highest power of 2 (e.g. reflective padding), alleviating boundary effects and yielding analyticity w/ fast decay.