Pspicker

- Fixed bugs(?) in amplitude calculation

- Updated PAZ to have separate initialization routines for pole-zero gain and
bandpass gain.

- Updated parameter file format:

yaml
channel_parameters:
compZ: 'Z3'
compN: 'N1Y'
compE: 'E2X'
compH: 'HF'
S_write_cmp: 'N'
P_write_cmp: 'Z'
P_write_phase: 'Pg'
S_write_phase: 'Sg'


becomes:

yaml
channel_parameters:
component_orientation_codes:
Z: 'Z3'
N: 'N1Y'
E: 'E2X'
H: 'HF'
write_components_phases:
S: ['N', 'Sg']
P: ['Z', 'Pg']


yaml

station_parameters:
SPOBS:
P_comp: 'Z'
S_comp: 'ZNE'
energy_frequency_band: [3, 30]
energy_window: 20 What does this really do?


becomes

yaml

station_parameters:
SPOBS:
picking_components
P: 'Z'
S: 'ZNE'
SNR_energy:
frequency_band: [3, 30]
window: 20 What does this really do?



- Got rid of sampling_rate attribute, a relic of C Baillard's code
which doesn't belong in a Pole-Zero

- Changed behavior of setting input_unit to an unknown value: previously
accepted the name, now keeps the old name if it can't convert units (was
a problem when saving unknown units to json_pz files, which forced units to
m/s, even if they were Pascals)

- Changed fast_kurtosis to reduce edge effect without throwing away first
win_samples points, by calculating using a win_secs length buffer of repeating
first second of data_

- Updated pyyaml and fixed some pathlib errors

- json_pz files now use m/s as default units
- Improved PAZ class
- Sorts input files by filename within each day

cleaned up parameter file reading, fixed log_level use in run_many()

- Parameter file changes:

- replaced ``kurt_frequency_band(s)``, ``kurt_window_length(s)``
and ``kurt_extrema_smoothing(s)`` to
``kurtosis:{frequency_bands, window_lengths, extrema_smoothings``
- renamed ``n_extrema`` to ``max_candidates``

- Changed some internal kurtosis code to simplify and better align with
Matlab code outputs
- Fixed a (probably unimportant) bug that kept kurtosis gradients from
reaching zero
- added a prepend and append to second np.diff in kurtosis._loca_ext(),
to retain array length and index alignment
- simplified smooth_filter, eliminating a bug that put zero in the first index