BUG/ENH: Improvements to step_until_intersect and bug fix for field_line_trace

[rstoneback]

A user ran into a calculation failure when the number of steps allowed for field_line_trace was large enough that the integration terminated early, resulting in downstream calculation errors. Corrected this by filtering outputs after the minimum distance to the desired termination height is reached. Downside: the extra calls to the ECEF/Geodetic conversion slow things down. Conversion to Fortran is worth it.

While investigating this issue, a problem with step_until_intersect was found and corrected. The method now does a better job searching for the closest distance. More accurate and more robust to changes in parameters. A tolerance keyword, 'tol', was added.

Corrected a small pysat version bug.

To be completed:

• Unit tests validating robustness over tol, step_size, etc.
• Add option to step_until_intersect for first step. This needs to go along with changes to the scalar mapping code to pass those variables on through.
• Create python/fortran ECEF/Geodetic method to improve runtimes (Created Issue #20)

[rstoneback]

Addresses issue #16

[rstoneback]

Test failures are related to some configuration issues on Travis CI. The code itself is fine.

[rstoneback]

Review requested from @bharding512

[rstoneback]

Some of the features here have been supplanted by the next pull. Next pull needs a bit of documentation work but the underlying code changes are good to go.

[rstoneback]

No rush as I've already delivered results to ICON.

[rstoneback]

Closed per feedback and notes in pull #26