do not force float64 dtype on assignment to Node, Node2D

[grlee77]

This PR removes the explicit casts to float64 from the wavelet packets code, enabling packets with the other dtypes supported by the DWT (float32, complex64 and complex128).

[grlee77]

Actually, on closer look, the dtype tests used here already pass on current master. It is only on assigning data to nodes that things were being forced to np.float64. Let me update the release note text and tests to reflect that before reviewing this one.

[codecov-io]

Codecov Report

Merging #392 into master will decrease coverage by 0.05%.
The diff coverage is 80%.

@@            Coverage Diff             @@
##           master     #392      +/-   ##
==========================================
- Coverage    84.5%   84.45%   -0.06%     
==========================================
  Files          22       22              
  Lines        3595     3596       +1     
  Branches      624      625       +1     
==========================================
- Hits         3038     3037       -1     
- Misses        489      490       +1     
- Partials       68       69       +1

Impacted Files	Coverage Δ
pywt/_wavelet_packets.py	77.57% <80%> (-0.56%)	⬇️

---

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 702d519...58cd9ce. Read the comment docs.

[grlee77]

Okay, this is good to review now. I have changed the label to bug since transforms in the other dtypes did work before. It is just assigning data to a node that caused the cast to float64. Only the test added in the last commit here fails on current master. The rest of the new tests pass on current master as well, but it doesn't hurt to test these other dtypes explicitly.

[rgommers]

LGTM, merged. Thanks @grlee77