Fix the leading dot notation in inserts

[zchenb]

Fixes #4958

[msullivan]

The problem is that while we wanted to disallow these leading dots in the expression itself, we do now allow them in defaults

[zchenb]

So I am not sure whether the following would make sense:

Given the schema

type A { required property  b := "hello"; required property  c := .b; required a : str; required d : str}

The following two represent potentially good insert queries:

1. Inserting with properties set to defaults:

insert A { a := .b, d := .c};

2. Inserting with properties set to another property:

insert A {a := "world", d := .a}

Both queries are currently allowed.

Obviously, we would like to ban queries like insert A { a := .a}, but it is currently unclear to me how to describe this subset. As obviously, by the same logic, insert A {a := .d, d := .a} is not allowed but insert A { a := .d, d := .c} is allowed.

[zchenb]

The current semantics seems to be that, all properties have defaults being the empty set unless otherwise specified. Inserts are really inserting the object after applying the insert shape to the default object. That is, insert A {S} is really inserting default(A) {S}.

[zchenb]

This PR is ready for merge!

[msullivan]

Looks good! Thanks.