Add preliminary Travis-CI support

[Echelon9]

This PR adds preliminary support for continuous integration, specifically via the Travis-CI platform.

Travis-CI is a cloud-hosted continuous integration platform, with good quality GitHub integration and free for open source projects.

How to enable:

• Individual developers who forked pypredict can enable Travis-CI on their own GitHub-hosted fork, via Travis-CI's GitHub integration. This allows each developer's branches to be run through CI.
• @nsat can enable Travis-CI for this upstream repo, turning on automated CI testing of all incoming PRs <branches> and master branch (with a potential status badge in README.md)

Comments:

• At the moment, the singular unit test fails regardless of whether run locally or on a cloud-based vendor. There are two known issues I've identified:

  • [x] (EASY) QTH format supplied in the unit test differs from that expected by the pypredict API. (https://github.com/nsat/pypredict/pull/30)
  • [ ] (MEDIUM) Sample TLE data supplied in the unit test is treated by pypredict as too stale in 2020. There's a debug override to address this in predict.c, however a better solution should exist.

• Really, pypredict should move to a modern testrunner framework e.g. pytest and pytest-runner.

Neither of these two comments need to be fixed before this support lands IMHO, as they can be addressed later. In fact, having this skeleton in place provides a good base to build upon and provides confidence in future refactoring efforts.

If the testsuite is known failing at the moment, how do I know a working testsuite will run green? Good question. I quickly hacked around the above two problems which returned green on Python 2.6 and Python 2.7.

If anyone is still unconvinced of the benefits of CI, here is a short set of reasons why.

[Echelon9]

I've rebased this PR on top of the master branch, and confirmed that the test suite runs still on Travis-CI infrastructure.

There remains a known test failure, whether run locally or on CI, due to above noted TLE staleness issue:

$ ./test.py

test_transits_are_truncated_if_the_overlap_the_start_or_end_times (__main__.TestPredict) ... ERROR

======================================================================

ERROR: test_transits_are_truncated_if_the_overlap_the_start_or_end_times (__main__.TestPredict)

----------------------------------------------------------------------

Traceback (most recent call last):

  File "./test.py", line 32, in test_transits_are_truncated_if_the_overlap_the_start_or_end_times

    next_transit = next(predict.transits(tle, qth, ending_after=at))

  File "/home/travis/build/Echelon9/pypredict/predict.py", line 66, in transits

    transit = quick_predict(tle, ts, qth)

  File "/home/travis/build/Echelon9/pypredict/predict.py", line 19, in quick_predict

    return _quick_predict(tle, ts, qth)

PredictException: time Wed 14Jan15 02:27:36 too far from present

----------------------------------------------------------------------

Ran 1 test in 0.002s

[miketwo]

I've addressed your second task with my PR. Locally, I used your test setup to develop the solution, so thank you for that.

[nick-pascucci-spire]

Hi, thanks for your pull request! We have recently added new maintainers to this project, including myself, and are addressing the outstanding PRs.

I've integrated GitHub Actions CI in #53 which is substantially similar to what you've proposed here, so I'm going to close this PR out.

[Echelon9]

No worries. Good to see that the project is under continued maintenance. All the best.