Address review feedback on split raw data storage (PR #8219)

[Copilot]

Applies fixes for the review thread on #8219, incorporating @astafan8's follow-up on the Copilot AI review comments.

Library helpers no longer print to stdout

purge_orphaned_datasets and cleanup_datasets are split into logging-only core helpers plus thin public callers that own user-facing output — keeping programmatic use quiet while preserving console feedback.

def _purge_orphaned_datasets(db_path, *, dry_run=True) -> PurgeResult:
    ...  # core logic, log.info(...) only

def purge_orphaned_datasets(db_path, *, dry_run=True) -> PurgeResult:
    result = _purge_orphaned_datasets(db_path, dry_run=dry_run)
    print(f"Found {result.total_datasets_with_raw_data} datasets ...")
    return result

Harden DROP TABLE

remove_dataset_from_db now runs _validate_table_name(result_table_name) before interpolating the name into DROP TABLE, guarding against dropping an unintended table if the stored value is corrupt.

Docs

• dataset_design.rst: corrected module path to qcodes.dataset._raw_data_storage.
• introduction.rst: removed the duplicated Database-notebook pointer.
• Fixed update_raw_data_paths docstring typo ("old paths3" → "old paths").

Out of scope

Two Code Review notes on pre-existing #8219 code were left untouched: the get_shaped_parameter_data_for_one_paramtree import (false positive — defined in queries.py) and a docstring spacing nit, neither part of the linked thread.