Tune wavelet transform

[anderskaestner]

Look into how the wavelet transform can be optimized for speed. Duplicates issue #38.

• [ ] Set up a unit test with benchmark part
• [ ] Make smart copying of the data to filter vector, remove step 2.
• [ ] Process in local blocks
• [x] Add STL threading
• [ ] Check if SSE instructions can help
• [x] Create python binding?

[anderskaestner]

The fftw3 library is not thread-safe only the execute functions are. The initialization needs guarded regions.