Bta-lib

- Indicators:
- `sar`, `sarext`
- Added missing check conditions in `ht_trendmode`
- Improved auto-docstring generation

- Indicators:
- `ht_dcperiod`, `ht_dcphase`, `ht_phasor`, `ht_sine`, `ht_trendline`,
`ht_trendmode`
- Allow indicators to declare numerically than less inputs than declared can
be accepted with `allowinputs=x`
- Refactor `mama` in terms of `ht_phasor`
- Add `_applymulti` primitive to allow an apply with multiple return results

- Indicators:
- `mama`
- Refactoring internal param names, fixing reduction op and other cosmetics
- Refactor obv as a function of _period and _setval
- Re-locate cumXXX and map as standard ops and not as reduction ops
- Add `_talib_class` classmethod fo ta-lib compatibility actions needed
before the instance is created
- API call to regenerate inputs
- New SEED_ZERO, SEED_ZFILL for _ewm
- New behavior for `_ewm` in which if alpha is passed but no `span` is
present, the actual minimum period of the calling lines is applied
- Added calls `mask` and `where`
- Correct from 0.9.6 that the dtype of the series was lost, which caused
problems with wrapped calls which actually check the dtype for example for
"boolean" like mask and where
- Internal semantics improved for the `__call__` notation and line `clone`
- Fix reduction operation minimum period
- Remove unneeded overlap parameter for standard operations and rolling
functions
- Added `_setval` method to set a value/range from the minimum period
- Extended `_period` to be able to set a value in the period which is
added/removed
- Refactor `obv` and others to use `_period`

- Indicators:
- `mavp`
- Make scipy.signal an optional requirement with a run-time import and
default to use _mean_exp if not available
- Refactor _mean_exp to use prev local variable and let it take an empty beta
(1-alpha)
- Report right attribute which was not found in Linesholder
- Add _minperiodize method which scans and adapts hidden series in args,
kwargs to minperiod
- Add line method which allow a dot notation period increase
- Add generic _apply method to series which mimics hidden _apply in "ewm" but
with args and kwargs, supporting raw for ndarrays
- Refactor standard_op to use minperiodize and some renaming
- Refactor reduction operations to use minperiodize and consider only the
minperiod
- Use SEED_AVG (0) enum as default seed instead of non-obvious False
- Add SEED_NONE logic to multifunc window operations
- Add reduction operation "unique" and adjust minperiod of series hidden in
kwargs in apply

- Indicators:
- `correl`
- Remove name setting because it may overwrite external inputs and change
macro-like extraction to fix name, adapting test framework
- Refactor crossover/up/down family, removing _type, simplifiyng formula and
adding a _strict parameter to switch between 2 bar crossover (strict) and a
crossover over multiple bars (non-strict)
- Macro-like additions: _MPSERIES, _SETVAL and _MPSETVAL to get series
relative to minperiod and set values (raw and relative to minperiod)
- Make "series" and "index" properties return only the raw underlying series
and index, to conform to public API to retrieve series
- Re-adapt test framework to series macro-like API re-change
- Remove old pandas macro module and replace with utils
- Refactor obv to use new _MPSETVAL macro-like function

- Indicators:
- `linearreg`, `linearreg_angle`, `linearreg_intercept`, `linearreg_slope`,
`tsf`
- Changed inputs_override/outputs_override to be the default behavior when
defining inputs in an indicator
- Added inputs_extend/outputs_extend to support a partial input definition
inheriting from the base
- Refactor all indicators using overrides to new ruling
- Add superclass _talib_ compat flag and refactor current _talib_ users to
consume it
- Add _SERIES macro to work with the underlying series
- Convert maxindex/minindex to work with _SERIES
- Convert all math indicators to "apply" instead of using series
- Make "series" and "index" properties return only the [_minperiod:] slice
and adapt test framework to change
- Refactor for logic of indicator auto-registration
- Use class name for parameterif the parameter is in the indicator hierarchy