Encode doubles as Double, do bounds checks, and add support for NaN and Infinity values

[drvirgilio]

See discussion in dhall-lang issue #252. I'm not sure exactly on where the parser for "[-]Infinity" should go so I kind of guessed.

[Gabriella439]

Looks great! I'll merge this once the upstream change to the standard is merged

[drvirgilio]

I merged the changes to the standard so I think this is ready to be merged.

[f-f]

@drvirgilio LGTM 👍
Could we add some tests that exercise the new keywords and conversions?

[Gabriella439]

@f-f: I'll go ahead and merge this now and add tests later, because I need to sync this repository with the dhall-lang test suite first. The other reason why is that I would like to add the tests to dhall-lang first and then mirror them to this repository rather than the other way around

[Gabriella439]

Actually, there are test failures on this branch that are being hidden by a CI misconfiguration on my part. I'm going to fix that first in a separate pull request and then fix the tests on this branch

[Gabriella439]

I have a pull request up against this branch with fixes to the test suite: drvirgilio#1

[drvirgilio]

Finally got around to making a few changes:

1. Added tests which check a few boundary conditions such as the point where Integer/toDouble rounds up to Infinity.
2. Found a workaround for fromInteger's weird rounding behavior by using read . show.
3. Made the changes suggested by @Gabriel439

[Gabriella439]

@drvirgilio: Thanks for doing this! 🙂