Statically verify that constant-valued text args to commands refer to valid entities

[kostmo]

In a .sw program, when we encounter a line:

place "foo";

given the context of a scenario, we should be able to check statically whether "foo" refers to an existing entity.

Entities are defined both globally and on a per-scenario basis.

• If we had a distribution format for scenarios (#915), the IDE could be made aware of any scenario-specific entity definitions when parsing any associated .sw files.
• When .sw programs are parsed at scenario-load time (e.g. with --autoplay or --run), could return immediate failure on incorrect entity name arguments instead of deferring failures to runtime.

Also related: #1301

[byorgey]

One slight practical difficulty is that we do not currently pass along the set of existing entities during typechecking. An alternative idea would be to check this in a separate pass, after typechecking has succeeded --- traverse through the term just looking for literal text arguments to functions that expect entities, and make sure they are known.

If we ever do #455 it would certainly change what this looks like, but that need not stop us implementing this useful (and hopefully not too difficult) check in the meantime.