Process schema files in a deterministic order

[msullivan]

Currently I think we concatenate the schema files in whatever order the OS lists them to us. I think we should sort the list first, so that when we hit server bugs that depend on ordering, it happens deterministically.

[msullivan]

Prompted by https://github.com/edgedb/edgedb/issues/6161, which I couldn't reproduce directly from the provided schema directory because of this nondeterminism