maint: add pyproject to allow installing from empty environment

[steven-murray]

This just adds a simple pyproject.toml that makes it possible to install this package into an env that doesn't already have numpy installed.

[coveralls]

Coverage remained the same at 42.65% when pulling e0c00ec on pyproject into 63110c8 on master.

[plaplant]

@steven-murray thanks for the change! I think we also want to move from Travis to GitHub Actions for the CI on this repo. I can set that up as part of this PR.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@63110c8). Click here to learn what that means.
The diff coverage is n/a.

@@            Coverage Diff            @@
##             master      #61   +/-   ##
=========================================
  Coverage          ?   47.87%           
=========================================
  Files             ?       17           
  Lines             ?     2258           
  Branches          ?        0           
=========================================
  Hits              ?     1081           
  Misses            ?     1177           
  Partials          ?        0           

Flag	Coverage Δ
unittests	47.87% <0.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 63110c8...29c4f73. Read the comment docs.

[plaplant]

As a part of this PR, I've moved the CI builds from Travis to GitHub Actions, moved the coverage reporting from Coveralls to CodeCov, and migrated the test suite from unittest to pytest. These changes bring the repo more in line with others across the organization.

As part of the testing, I have removed the builds on python2.7, and added testing for python3.7, 3.8, and 3.9. I have not explicitly broken compatibility with python2, but I don't think we need to regularly test backwards compatibility any longer now that it has reached end-of-life.

In addition to these changes, I've been more explicit about numpy datatypes, in keeping with the recent deprecation warnings introduced in v1.20. This should hopefully future-proof us a bit.

[steven-murray]

Nice updates, @plaplant :-) Looks ready to go!