Tomli

- Improve
- Make bundling easier by using relative imports internally and adding license and copyright notice to source files.
- Make error messages more uniform
- Raise a friendly `TypeError` for wrong file mode
- Allow `parse_float` to return objects having the `append` attr
- Eagerly raise an error if `parse_float` returns an illegal type
- Packaging
- Move from `pytest` testing framework to `unittest` and remove `python-dateutil` test dependency.
Tests now only require Python interpreter.

- Removed
- Python 3.6 support
- Support for text file objects as `load` input. Use binary file objects instead.
- First argument of `load` and `loads` can no longer be passed by keyword.
- Fixed
- Allow lower case "t" and "z" in datetimes
- Improved
- Raise an error when dotted keys define values outside the "current table".
Technically speaking TOML v1.0.0 does allow such assignments
but that isn't intended by specification writers,
and will change in a future specification version (see the [pull request](https://github.com/toml-lang/toml/pull/848)).

- Fixed
- Backport: Allow lower case "t" and "z" in datetimes

- Fixed
- Illegal characters in error messages were surrounded by two pairs of quotation marks
- Improved
- `TOMLDecodeError.__module__` is now the public import path (`tomli`) instead of private import path (`tomli._parser`)
- Eliminated an import cycle when `typing.TYPE_CHECKING` is `True`.
This allows `sphinx-autodoc-typehints` to resolve type annotations.

- Fixed
- Raise an error if a carriage return (without a following line feed) is found in a multi-line basic string
- Type annotations
- Type annotate `load` input as `typing.BinaryIO` only to discourage use of deprecated text file objects.
- Packaging
- Remove legacy `setup.py` from PyPI source distribution.
If you're a packager and absolutely need this file, please create an issue.

- Deprecated
- Text file objects as input to `load`.
Binary file objects should be used instead to avoid opening a TOML file with universal newlines or with an encoding other than UTF-8.