GPU & CPU acceleration; caching

[OverLordGoldDragon]

Dramatically speed up compute and reduce memory usage

FEATURES:

• GPU acceleration & multi-thread CPU support for all forward transforms (cwt, stft, ssq_cwt, ssq_stft); see Performance guide
• ssqueezepy.FFT, supporting single- & multi-threaded CPU execution, and GPU execution, optionally via pyfftw
• dtype='float32' and 'float64' support for cwt, stft, ssq_cwt, ssq_stft, Wavelet
• Wavelet.Psih(scale=, N=) will store the computed wavelet(s) and, if subsequent calls have identical scale and N, will return it directly without recomputing (significant speedup).

BREAKING:

• Dependency added: ftz
• Default downsample: 3 -> 4 in utils.cwt_utils.make_scales
• EPS deprecated in favor of EPS32 & EPS64 for respective precisions
• ssq_cwt(flipud=True) default now returns Tx = np.flipud(Tx) relative to previous versions
• ssq_cwt & ssq_stft number of variables returned now depend on get_w, get_dWx parameters; see docstrings
• dtype defaults to 'float32' (can change via configs.ini); neither cwt nor stft, for most applications, require extreme precision like filters do, so defaults should prioritize compute

FIXES:

• ssq_stft would still default n_fft = len(x); defaulter line removed, delegated to stft.
• ssqueezing: improperly handled return of infer_scaletype
• ssqueezing: _get_center_frequency computed at N instead of p2up(N) with padtype != None

MISC:

• configs.ini: added new configurable defaults
• ssqueezing.ssqueeze: added padtype arg (see FIXES)
• ssqueezing.ssqueeze & ssq_cwt: added find_closest_parallel arg (see its docstring)
• utils.cwt_utils: find_downsampling_scale added argument N
• visuals.imshow: removed default 'interpolation' = 'none'
• Added # Arguments: docstring to Wavelet
• cwt significantly sped up: 1) per Wavelet reuse; 2) rid of ifftshift and *pn (they undo each other); 3) eliminated redundant allocation in vectorized

NOTES:

• Undocumented changes; skimming docstrings / source code should suffice for most purposes

[codecov-io]

Codecov Report

:exclamation: No coverage uploaded for pull request base (master@1ac7ed2). Click here to learn what that means. The diff coverage is n/a.

@@            Coverage Diff            @@
##             master      #32   +/-   ##
=========================================
  Coverage          ?   92.77%           
=========================================
  Files             ?       17           
  Lines             ?     3377           
  Branches          ?        0           
=========================================
  Hits              ?     3133           
  Misses            ?      244           
  Partials          ?        0           

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 1ac7ed2...4cdcee2. Read the comment docs.