Fortranformat

- Feature: [Issue 6](https://github.com/brendanarnold/py-fortranformat/issues/6) Improve performance of reading and writing of `F` format by x10 - ZedThree
- BREAKING: Python 3.6 now required

- Bug: [Issue 28](https://github.com/brendanarnold/py-fortranformat/issues/28) Fix a bug where crash occurs on outputting infinity or NaN for a floating point number type.

- Feature: [Issue 25](https://github.com/brendanarnold/py-fortranformat/issues/25) Include a minimal test suite for use where resources are limited e.g. pipeline builds

- Bug: [Issue 21](https://github.com/brendanarnold/py-fortranformat/issues/21) Now outputs FORTRAN default values when `None` is passed.

- Bug: [Issue 15](https://github.com/brendanarnold/py-fortranformat/issues/15) Fix hanging when no suitable edit descriptor specified for `G_INPUT_TRIAL_EDS`. Now raises `ValueError`

- Bug: Fixed overflow behaviour in tests, in particular fixing the PROC_MAXINT behaviour
- Bug: [Issue 17](https://github.com/brendanarnold/py-fortranformat/issues/17) Properly outputs zero dp floating point numbers
- Migrated to development on Python 3 (non-test code still compatible with Python 2)
- Generally used more standard project structure as detailed at [https://docs.python-guide.org/writing/structure/](https://docs.python-guide.org/writing/structure/)