Add functionalities

[rohaquinlop]

This pull request introduces several improvements and refactors to the mock ScyllaDB implementation, focusing on more robust parsing and handling of primary keys, clustering orders, and lightweight transaction (LWT) conditions. It also enhances the parsing of CQL statements and improves the accuracy of table metadata representation. The most significant changes are grouped below:

Primary Key and Clustering Order Handling:

• Refactored the parsing and storage of primary key information to distinguish between partition and clustering keys, supporting both inline and composite definitions. Added support for storing and utilizing clustering order information for columns, and ensured this metadata is reflected in both table and column representations (mockylla/parser/create.py, mockylla/classes/metadata.py, mockylla/classes/state.py).

CQL Statement Parsing Improvements:

• Improved regular expressions for parsing CREATE TABLE, INSERT, SELECT, UPDATE, and DELETE statements to better support CQL syntax, including optional clauses and more flexible LIMIT expressions (mockylla/parser/__init__.py).

Lightweight Transaction (LWT) and Conditional Deletes:

• Refactored the DELETE statement handler to support conditional deletes using LWT (IF, IF EXISTS, IF NOT EXISTS), including proper result row generation and integration with materialized view rebuilding (mockylla/parser/delete.py).

Expired Row Purging:

• Added logic to purge expired rows from tables before returning data, ensuring returned results are up-to-date with TTL semantics (mockylla/classes/state.py).

Code Structure and Utility Improvements:

• Introduced utility functions for splitting and parsing CQL components, and reorganized code for clarity and maintainability, such as extracting top-level comma splits and normalizing WHERE clauses with parameters (mockylla/parser/create.py, mockylla/parser/delete.py).

These changes collectively improve the accuracy, maintainability, and feature completeness of the mock ScyllaDB engine, especially around schema handling and query parsing.