Remove field_identifier from keyed_element

[adonovan]

This change removes field_identifier from keyed_literal, since in the expressionT{k: v}, k may be:

• an expression, if T is a map, slice, or array type;
• a composite literal without an explicit type, if T is an array type;
• a field identifier, if T is a struct type.

There is no way to distinguish the first two cases without type information. Since expressions include identifiers, it's not necessary to add a special case for field_identifier, and it is incorrect when the identifier is actually a var or const.

Also, factor the grammar using commaSep.

Checklist:

• All tests pass in CI.
• There are sufficient tests for the new fix/feature.
• Grammar rules have not been renamed unless absolutely necessary.
• The conflicts section hasn't grown too much.
• The parser size hasn't grown too much (check the value of STATE_COUNT in src/parser.c).

[aryx]

The tests need to be updated

[adonovan]

The tests need to be updated

Of course. I did update them, and then all the corpus text files got reformatted unnecessarily, and I thought I'd finally found a way to avoid that, but it turns out it was by accidentally removing the tests from the PR. Ugh, I don't know what's going on with me this week.

PTAL.