Fix --field not applying pending contracts

[yannham]

Closes #1774

The --field argument has been added in 1.3 to act only on a specific field of the configuration. However, pending contracts (contracts attached to the field by metadata annotations or propagated through merging) weren't properly applied, leading to illegal values being accepted.

This commit fixes the issue by adding the missing contract application step. A snapshot test is added as a regression test. While it's not ideal (we don't encode in the test what is the expected result), it's the best we can do currently for CLI-only tests.

I plan to publish a minor release with this patch, as silently not applying contract is serious enough.

[vkleen]

This seems like a good quick fix. Long term, I wonder if it wouldn't be better to encapsulate the "extract value out of field" operation so that we'd have a harder time forgetting to do that. Now, there are quite a few places in the code base that need to call RuntimeContract::apply_all and the whole setup feels quite fragile.

[yannham]

This is a good point. A simple helper first, like value_with_ctrs(&self) -> Option<RichTerm> would probably go a long way, at least for deduplication. For correctness, we might end up making the value field private, even if general I don't like accessors proliferation when the language provide a built in way to access struct fields, but I guess encapsulation (for correctness) is a valid motivation.