Require record fields to be simple identifiers

[shonfeder]

When a :: appears in a quint identifiers, I am referring to that identifier as
"qualified". E.g., Foo::bar is qualified, but _ding is not.

This takes care of one problem identified in #1081 by introducing a syntax error
a record field is a qualified identifier.

Along the way, the PR also gives further illustration of how we can address #17
by using error grammar rules. (There was already precedent for this in the
errorEq rule checking for the misuse of = for equality). This shows how we
can move stronger syntactic checks into the grammar without sacrificing the
quality of our error messages.

E.g., errors for this new rule look like so:

>>> type r = {a: int, b::c::d: str}
syntax error: error: [QNT008] Identifiers in a record cannot be qualified with '::'. Found b::c::d.
type r = {a: int, b::c::d: str}
                         ^

To enable this improvement in the grammar, we had to refactor the lexical rule
for IDENTIFIER: without splitting this into a lexical rule for the
unqualified case and a grammar rule for the qualified, we end up with an
ambiguous lexical rule that makes parsing fail.

One incidental change is added to make the reference field in our error type
optional. This is implied by the fact that parser errors are errors that should
be represented with this structure, but we may not have a reference for them yet.

• Tests added for any new code
• [-] Documentation added for any new functionality
• Entries added to the respective CHANGELOG.md for any new functionality
• [-] Feature table on README.md updated for any listed functionality

[shonfeder]

The last three commits are generated/mechanical!

[thpani]

Looks great!

I have an alternative suggestion to making the error type fields optional; not sure myself if I prefer one over the other.