IO update for RTOUT frequency and metadata

[kafitzgerald]

TYPE: bug fix + "enhancement"

KEYWORDS: IO, medium range, retrospective, output frequency

SOURCE: Katelyn, NCAR

DESCRIPTION OF CHANGES: Similar to the retrospective model output configuration, a reduced output frequency was desired for the terrain routing grids in the medium range forecast outputs (every third file).

This was implemented, the number of valid_times in the output files was updated, and the timing issue (outputs should occur for hours 3,6,9... and not 1,4,7... for hourly outputs) was addressed.

ISSUE: Closes #477

TESTS CONDUCTED: Full test suite.

NOTES: This is really just an update to an existing hack. In the longer term we should think about a better way to specify output frequencies for different file types.

Checklist

Merging the PR depends on following checklist being completed. Add X between each of the square brackets if they are completed in the PR itself. If a bullet is not relevant to you, please comment on why below the bullet.

• [X] Closes issue #477
• [ ] Tests added (unit tests and/or regression/integration tests)
• [X] Backwards compatible
• [ ] Requires new files? Nope
• [ ] Fully documented
• [ ] Short description in the Development section of NEWS.md

[rcabell]

Travis CI tests are failing because they don't correctly target the base branch of the PR (v5.2.0 in this case) -- Travis always targets master.

The tests on Cheyenne do pass, because prtest is smart enough to determine what the base is.