Bruges

- Fixed small issue with the Klauder wavelet not working when `autocorrelation` was `False`.
- Changed the names of the `convolve` arguments to `arr` and `v` to better match NumPy.
- The version number is now correctly stored in the `__version__` attribute.
- Added a User Guide notebook on [making wavelets](https://code.agilescientific.com/bruges/userguide/Making_wavelets.html).
- Added a User Guide notebook on [seismic attributes](https://code.agilescientific.com/bruges/userguide/Seismic_attributes.html).

- Fixed [issue 92](https://github.com/agilescientific/bruges/issues/92): the Ormsby, Ormsby FFT and Klauder wavelets now behave as expected when 2D arrays (lists of frequency parameters, essentially) are passed. A filter bank (array of wavelets) is returned.
- Fixed [issue 93](https://github.com/agilescientific/bruges/issues/93): the Sinc, Cosine and Gabor wavelets now behave as expected when `sym` is passed.
- `reflection.convolve()` now works properly when the axis is specified, and a verbose option was added to show the input and output array shapes (helps with troubleshooting).

- **Breaking change:** Mode is now `'same'` by default on reflectivity functions. If you were assuming `mode='valid'` you should change your code.
- `reflection.reflectivity()` now works properly on 2D panels of Vp, Vs, and rho.
- `reflection.acoustic_reflectivity()` now optionally accepts *only* Vp (or impedance) or Rho. You should pass both if you have them, but if you only pass one, reflectivity will be computed from that alone.
- `reflection.convolve()` now works properly on 2D and 3D reflectivity series, and even works with 2D wavelet banks (provided the bank is shorter in time than the model, which it usually will be).
- `reflection.elastic_impedance()` now works properly on 2D panels of Vp, Vs, and rho.
- We have started adding better documentation; check it out at [code.agilescientific.com/bruges](https://code.agilescientific.com/bruges).
- `models.wedge()` now behaves as expected if you pass a NumPy array of `int` values to `strat`.

- Fixed a critical bug in 0.5.0 which prevented it from importing.

- In all wavelets, `sym` is now `True` by default.
- In all wavelets, `return_t` is now `True` by default. you can still set it to false but...
- **Future breaking change:** In the next minor release, `return_t` will always be `True`. **If you are setting it to `False`, please change your code.** We will continue to accept the argument for the time being, so if you're setting it to `True`, you can stop doing that, and leave it out of future code.
- **Breaking change:** I changed the names of several of the new discontinuity attributes to use the word 'discontinuity', instead of 'coherency', 'similarity', etc. This is in line with Art Barnes's classic paper "Too many seismic attributes". It was too confusing.
- `bruges.filters.wavelets.rotate_phase()`, deprecated in v0.4.1, has now gone away and is available in improved form as `bruges.filters.rotate_phase()`.
- Allowed discontinuity attributes to work properly on 2D data.
- The docs have been substantially overhauled; we hope you notice an improvement.

- **Breaking change:** `bruges.attribute.similarity` (n.b. now renamed `discontinuity` in version 0.4.6) has a new interface and computes its result(s) differently. I added several seismic coherence attributes as options to `bruges.attribute.similarity`. Methods include Marfurt coherence, that of Gersztenkorn, A., and K. J. Marfurt, 1999, and gradient structure tensor coherence (T Randen et al., 2000, 70th Annual International Meeting, SEG, Expanded Abstracts, 668-671.). Choose the method with the 'kind' argument. These functions were written by Joe Kington and published in The Leading Edge in the June 2015 issue.
- Complex trace attributes `instantaneous_amplitude` (envelope), `instantaneous_phase` and `instantaneous_frequency`, along with the `quadrature`, were added to `bruges.attribute`.
- Fixed a bug in `bruges.models.wedge` that led to some wedge models not having the correct total size in the 'depth' dimension (e.g. sometimes being off by one pixel.)
- Now tested on Python 3.8, 3.9 (should have been working, just catching up!). Removing Python 3.4 and 3.5 from testing.
- Coming soon: dynamic-time-warp-assisted panel interpolation.