Loosen restriction on optional arguments.

[dustinswales]

When optional arguments were added back into the CCPP as part of #189, all variables that were conditionally allocated on the host side were required to be defined as optional within the schemes. This restriction required host-model logic to be implicitly built into the schemes and metadata. For example, scheme variables that were optional in the context of the host model, but always required within the context scheme, still needed to be defined as optional within the scheme. This resulted in schemes having an arbitrary/confusing mis-match of required/optional arguments.

Alongside this PR is a PR into the CCPP Framework that removes this restriction.
The changes contained within here revert nearly all of the optional argument changes introduced by #189 to the primary schemes.
The UFS interstitial schemes are untouched in this PR, as these schemes are designed to flexible (optional arg heavy) in order to couple to many primary schemes.

@climbfuji

[climbfuji]

This is too much to review line by line. We will need to rely on the UFS regression testing. As long as all tests pass with b4b identical results, this is good. Thanks for all your efforts!

[dustinswales]

This is too much to review line by line. We will need to rely on the UFS regression testing. As long as all tests pass with b4b identical results, this is good. Thanks for all your efforts!

@climbfuji Agreed.
I found a small metadata bug when running the full tests overnight. Re running now.