Reaction-plane-fit

Added

- Additional documentation. See: `c49fa753201250734f6dbc4bdce6a64959a277cb`.
- Integration test for differential signal fit. See: `efcfc40fdf662d5fb0297ed1d02c2f6b538e2485`.

Changed

- Switch emphasis to fitting **before** normalizing by the number of triggers, including updating default
arguments and documentation. See: `17a5d24397c313c3c5a39964e9a8bd471161153e`,
`e04a1f3b1736c8baef41872680bd175f3dadd977`, `73f5f0d8dcbf7f86f1742e57279da43fd62faa1f`.
- When reading the RP fit object from YAML, setup the entire fit object. See: `1b38175ae421729a1933bae63bcc6652262927d3`.
- Improve Minuit fit strategy. See: `d8dfb49d0599f564597461270b73bb67d1818bdf`.
- Rename `_fit` to `cost_func` for clarity - it represents the cost function, not the fit function. And it's
sometimes useful to access from outside the object (such as when calculating effective chi squared). See:
`3dc750119be1f52f05134b11f24883b2e5c73b07`.
- Make the Minuit verbosity a user settable option. See: `22e67c85b630ef09a3105bcee4587e8d7091b293`.
- Allow a user to skip HESSE when necessary. This should be rare, but it is required at times. See:
`98104086ca505a25778c75a0c63053085a836368`.
- Updated the signal differential fit procedure. The inclusive case is just treated as a background only fit.
This is required because adding coupling gaussian distributions is not a straightforward process (unlike
if they were uncoupled). Consequently, extracting the inclusive width relies on fitting to information already
included in the RP fit (included via the differential signal fit components), which means that the covariance of
the fit is neglected. Use care when using this fit approach! (A warning is thrown when using it). See:
`7ba221d8effcda48d3a5245776aefcb971d7748e`.
- Sample data now acts as if it is the raw correlation rather than scaled by the number of triggers. See:
`5c613d2f5d21882fe9d61d6d0067bd602cb74ff2`, `9484b457012f6647283e3fc1ebf43a719a0fd7e5`,
`5c613d2f5d21882fe9d61d6d0067bd602cb74ff2`, and `7db8ea4e4201879c609c0cebeb211ea67570e7ba`.
- Improve testing procedure by loading stored fits from YAML rather than defining them by hand in the test
file. This makes any future updates much easier. See: `5c613d2f5d21882fe9d61d6d0067bd602cb74ff2`.

Fixed

- Log likelihood fits by updating the underlying `pachyderm.fit` package. See: `33bdcb41cd0f2dcc6e76f8760b8e5eddd24456dd`.
- Minor typos in code, documentation. See: `150afbba7f065f54895f1bd30409c66df14990b9` and
`56af2759823cf2ff027c2b89d7047fdcb2bfc98f`.
- Create output directory when writing if necessary. See: `7ca7f01d12a897907427972c6a9f6402cba6c616`.
- Allow specification of `v_{1}` values. See: `ed037ba57f72dc24257dd6a2dbe1351e85e6c73d` and a fix in
`63cb49906eabc115f58b6847de756346b61b08c3`.
- Updated typing. See: `db79b4134f57ff76727356654489b9a823cb6b67`.
- Integration tests run successfully on TravisCI again. See: `330c3bc6a7f90f1a5e0773c03170a8b27b8cf4ca`.

Changed

- Added a method to create a full set of fit components for each EP orientation, regardless of which were used
for the fit. This allows for downstream code to not need to differentiate between the different cases for
most tasks, easing the user experience. See: Nearly all of the commits between 3.0 and 3.1.
- Use the `pachyderm.fit.BaseFitResult` for components in order to avoid needing to calculate unnecessary
values for component fit results. It's more natural. See: `bbbcac45c41490072f345b6aebbeb6f49d427a1d`.
- Updated the included test data for better test consistently with external sources. See:
`0bf5e1ceaede9cd578a20c12db81455f38326a2d`.
- Updated pre-commit hooks. See: `6d780545b5547f42c7c3a684f11234d069973c10`.

Fixed

- Warning about registering `pytest` marks. See: `8f6f5386fddea1e540a15de99ce8852dfc2319fd`.
- Removed final `probfit` dependence. See: `b94ea052246269e454f3df1c7afbb7ac96f35cec`.

Changed

- Move to using the pachyderm fit package for general fit functionality, cost functions, and fit results. This
move was instigated by the recognition that `probfit` cost functions didn't actually work properly with
pre-binned data. It also allows for the centralization of basic fit code to make it generally available.
Some code from this package was refactored into pachyderm. See: `9e62f6777afed290709b74d4ee18284d61c2bab9`
and `0f66cff8264ff06f5138d78ba0213f40a195dd5e`. Note that developed was handled in a private branch in this
repository, so some older commits have references to a fit package in the RPF package before it was moved to
pachyderm.
- No longer rely on `probfit`. All functionality and more is now provided in `pachyderm.fit`. Plus, the package
wasn't super actively maintained. See: `9e62f6777afed290709b74d4ee18284d61c2bab9`.
- Updated pre-commit hooks. See: `03844495a12f4269b14dbd4972c09fb203a88c7c`.

Fixed

- Log likelihood and chi squared cost functions properly take binned data into account. See:
`0f66cff8264ff06f5138d78ba0213f40a195dd5e`.

Added

- Enabled strict `mypy` checks. See: `1585c7398201319bac49f6881f678be3f9582ee5`.

Changed

- Improved typing. `341330a5e48f1526848de27ceaba4ee10499de67`.

Changed

- Refactor function error calculation so that it can used by other packages. See:
`04dff6cc7cbd1fd4ad1f2f26ecc528dfeebbe3f7`.

Fixed

- Enable fixing parameters via user arguments. See: `cf9de684245217d4c7583407940d5a3864b192b1`.
- Properly mark the package as typed. See: `fe43d5368bb8fdd1defa5dbff8a39ac45d2424a3`.
- Improved code formatting and typing information, and fixed typos.
- Fix `mypy` typing in third party packages by adding `zip_safe = False` in the `setup.py`. See:
`79eb4e821bed8d5290c18c779c2bce6c8a71e5c0`.

Added

- Calculate the correlation matrix in the fit result. See: `ee94cb80e774c38848f31592b8df765c333fba7a`.
- Read and write fit results with YAML, including integration tests. See:
`983a2519d9b882a34d6bf44be312b45a3e051316`.
- Added test for invalid fit arguments. See: `009a974dc038d79ad2890e9c3f935094a88cb3ee`.

Changed

- Modify the `ReactionPlaneFit.fit(...)` to return the minuit object for advanced use cases. The fit result
stores almost all of the minuit information, but it can be useful in limited circumstances. For now, it is
only used to check the correlation matrix calculation. See: `7a1efd908203abf1a1b8cb8d2f9d76afdfb1442d`.
- Initialize the component fit functions when defining the fit objects. This allows fit components to be
evaluated using stored fit results. See: `e2673f25e7fa952087bb45ab5e35204850138b69`.
- Updated typing information. See: `0f284c79ccfb6ee04715b33e4500b2c96832f7ff` and
`60af85240d84adf5e011ee0b4b2f7a0c8cef154e`.
- Updated documentation. See: `b94fc955ce15034490888ec75368853f61ad9244`.
- Updated pre-commit hooks with newer versions of the packages. See:
`6d54231e4707502fec9811c0507daa5a37358076`.