Welly

- Fixed [bug 208](https://github.com/agile-geoscience/welly/issues/208) which essentially meant `rename_alias` was always `False` when getting a well or project as a DataFrame.
- Fixed [bug 209](https://github.com/agile-geoscience/welly/issues/209) that prevented Well (and therefore Project) from exporting a DataFrame for some recent versions of `pandas`.

- `Curve.values` now returns a 1D array for a 1D curve.
- Added `Project.basis_range` to provide the min and max of all curves in the project.
- Added [a user guide page for `Project`](https://code.agilescientific.com/welly/userguide/projects.html).
- Fixed [bug 202](https://github.com/agile-geoscience/welly/issues/202) with curve indexing.
- Fixed [bug 206](https://github.com/agile-geoscience/welly/issues/206) that prevented quality tests from running on aliased curves.
- Fixed [bug 207](https://github.com/agile-geoscience/welly/issues/207) that was causing the quality table not to render correctly in some situations.

- Major change: Everything in `welly` is now much closer to `pandas`. Chiefly, `Curve` objects are now represented by wrapped `pandas.DataFrame` objects (note: not `Series` as you might expect, so they can be two-dimensional). They were previously subclassed NumPy `ndarray` objects, so while we've tried to preserve as much of the API as possible, expect some changes. Please let us know if there's something you miss, it can probably be implemented. Many thanks to the developers that made this happen, especially Patrick Reinhard and Wenting Xiong in the Netherlands.
- Major change: as previously indicated, the default behaviour is now to load the depth curve in its original units. That is: `welly` no longer converts everything to metres. Use `index='metres'` in `from_las()` to get the old behaviour.
- Major change: the `Well` object's header, `well.header`, is currently a large `pandas.DataFrame` containing everything from the LAS file's header. In the next minor release, we will restore something more like the original header object. We welcome opinions on how this should work.
- The `Curve` object should be instantiated with `index` instead of `basis`.
- You can now create a project with `welly.read_las('path/to/*.las')`. Note: this always gives you a project, even if it only contains a single well. You can get the single well from a path like `'data/myfile.las'` with a singleton assignment like `well, = welly.read_las('data/myfile.las')`.
- As previously indicated, dogleg severity is now given in units of degrees per course length.
- `kwargs` are passed to `lasio` in `read_las()`, `Well.from_las()` and `Project.from_las()`, so you can add things like `mnemonic_case='preserve'` or `ignore_header_errors=True`. See [the Lasio documentation](https://lasio.readthedocs.io/en/latest/) for more on these options.
- A new argument on `well.to_las()` allows you to control the case of the mnemonics in the output LAS file. The behaviour has always been to preserve the case in the data dictionary; choose 'upper', 'title' or 'lower' to change it.
- New docs! They are live at [code.agilescientific.com/welly](https://code.agilescientific.com/welly). Feedback welcome!

- No longer supporting versions of Python before 3.6.
- `Curve.top_and_tail()` has been implemented. It removes NaNs from the start and end — but not the middle — of a log.
- You can now optionally pass any of `start`, `stop` and `step` to `Well.unify_basis()`. These settings will override the basis you provide, or the basis that `welly` discovers using `Well.survey_basis()`. I added an example of using this to the `tutorial/02_Curves.iynb` tutorial notebook.
- Relatedly, if you pass any of `start`, `stop` and `step` to `Curve.to_basis()` it will _override_ the basis you give it, if you give it one.
- Welly now uses [`wellpathpy`](https://github.com/Zabamund/wellpathpy) to convert deviation data into a position log. The API has not changed, but position logs can now be calculated with the high and low tangential methods as well.
- Dogleg severity is still given in radians, but can be normalized per 'course length', where course length is a parameter you can pass. **Future warning:** from v0.5.0, dogleg severity will be passed in degrees and course length will be 30 by default.

- Fixed a bug that was preventing Alias names from appearing in the DataFrame view, `project.df()` and `well.df`. Updated the `Project` tutorial to reflect this.
- Fixed a bug that was preventing Aliases from applying properly to well plots.
- Improved the error you get fro `w.plot(tracks=[...])` if there are no curves to plot (e.g. if none of the names exist).

- Reorganized the `tutorials` a bit and made sure they all run as-is.
- A new `Location.from_petrel()` function accepts a Petrel `.dev` deviation file. It will extract the x and y location, and the KB, as well as the position log and/or deviation survey.
- `Curve.plot_2d()` now handles NaNs in the curve.
- The test functions now accept a `keys` argument to limit the number of items the tests will be applied to, or to order the appearance of curves in `qc_table_html`. For example, if you pass `keys=['GR']` then tests will only be run on `w.data['GR']`, regardless of what's in the `tests` dictionary. This was [issue 104](https://github.com/agile-geoscience/welly/issues/104).
- You can now pass a `pathlib.Path` object to `from_las`. Thank you to Kent Inverarity for implementing this feature.
- Added `XCOORD` and `YCOORD` as standard fields; these are read in as `location.x` and `location.y` repsectively.
- Added `Project.plot_map()` to make a quick (ugly) scatter plot from x and y location (whatever two field you provide from the `location` object).
- Added the `filter_wells_by_data()` method to `Project`, and deprecated `find_wells_with_curve()` and `find_wells_without_curve()`. You can make complex selections with this function, such as "give me all the wells that have at least two of RHOB, DTC or DTS".
- Added the recently added `index` argument (to preserve depth units) to `Project`.
- The LAS header items EKB and EGL are now captured as `ekb` and `egl` in the `w.location` object. KB and GL are captured as `kb` and `gl`.
- Thank you Miguel de la Varga for an update that allows a trajectory to have fewer than 3 points.
- Thank you DC Slagel for an update that ensures all well header fields are populated with valid types.