Make `CreateTable.location` / `AlterTable.location` make more sense

`CreateTable.location` has no documentation and is a tricky edge case, not always set even when SQL statement did set location.

https://docs.rs/sqlparser/latest/sqlparser/ast/struct.CreateTable.html#structfield.location

Since this took some time to figure out, here's what I've found for others also pondering its role:

- `CREATE EXTERNAL TABLE .. LOCATION 'foo'` sets `CreateTable.location`
- `CREATE TABLE .. LOCATION 'foo'` sets `CreateTable.hive_formats.location`
- `ALTER TABLE .. LOCATION` sets `AlterTable.location`
- `AlterTable.location` types are about Hive, and is thus a bit confusing for external tables: https://docs.rs/sqlparser/latest/sqlparser/ast/enum.Statement.html#variant.AlterTable.field.location
- alter vs create behaviors thus do not match

Maybe make Hive syntax use the top-level `location` field?

Start with the docs.rs definitions for CreateTable.location, CreateTable.hive_formats.location, and AlterTable.location, then trace the parser behavior for the CREATE EXTERNAL TABLE, CREATE TABLE, and ALTER TABLE LOCATION forms listed in the issue. Done means the location semantics are made consistent or clearly documented, with regression coverage for those cases.

Written by the indexing model from the issue text.