Fix 28 Implement SDP of MASS_V4

[NimaSarajpoor]

This PR is to address #28.

[gitnotebooks]

Review these changes at https://app.gitnotebooks.com/stumpy-dev/sliding_dot_product/pull/30

[NimaSarajpoor]

According to the following plot, challenger_sdp (blue) does not show a good performance. Note that the batch-processing part of "MASS_V4" has not been added yet to challenger_sdp. The following plot (most likely) confirms that we cannot just use Discrete-Cosine-Transform(DCT)-based SDP alone and we may need to add that batch processing piece, and use this algo for large n (>= 2^20).

./timing.py -timeout 1.0 -pmin 6 -pmax 20 pyfftw challenger pocketfft_r2c_c2r scipy_oaconvolve > timing.csv

[seanlaw]

Yeah, it looks pretty bad right now! Fingers crossed that things improve. Otherwise, we should/could reach out to the authors.

I find it curious that DCT is found in scipy.fft