Split migrations from the current executable

Now, the current code produces one binary that deals with everything, including MIGRATIONS.
Preserve the same logic, but produce another binary that deals with migrations.

1. Split into two binaries, one for migrations, the other remains the same.
2. The scheme uses one "schema.hcl"; split this into groups of multiple tables, context-based, e.g., everything starting with "XAGO_" should be "xago.hcl" containing the tables needed.
3. By default, the command runs with no flags and spits the SQL into stdout. This should be the default mode. ./binary -> outputs the SQL generated, ./binary --some-other-flag-with-no-name-now, it does the migration "--prod" for now.
4. Testing, using the current migration scheme, compare with the new one; the output should be identical.

Start by reading the current migration scheme and schema.hcl to understand how the existing binary produces and applies migrations. Separate the migration handling from the other executable, organize tables into context-based schema files such as xago.hcl, and verify that the default command writes SQL to stdout while the migration mode applies it. Compare the new output with the current scheme and require identical SQL.