Parsing with pglast instead of regex

Currently, parsing of schema dumps into various objects is achieved with the usage of regex. This is not maintainable on the long run.

[pglast](https://github.com/lelit/pglast) is a PostgreSQL Languages AST, it supports visitor patterns and streamlines the process of parsing of schema dumps into various objects.

Also, we should use this opportunity to make the project more structured and installable through the usual pypi repository.

No files, tests, or entry points are named. Start by locating the current regex-based schema-dump parsing and reviewing pglast's visitor patterns, then determine the packaging structure needed for PyPI installation. Done means the parsing approach and project installation are migrated as described without relying on the existing regex workflow.

Written by the indexing model from the issue text.