Tolerate missing additional data tables in code quality visualizations

[Jammy-Louie]

Problem

complexity_vs_test_gaps_bubble and code_quality_executive_dashboard load their additionalDataTables (the secondary tables declared in their specs) with raw pd.read_csv(...). When the notebook runs without a path supplied for one of those tables, the parameter keeps its in-package default (e.g. ../samples/test_gaps.csv), which does not resolve in the run environment, and the notebook aborts before it can render anything:

FileNotFoundError: [Errno 2] No such file or directory: '../samples/test_gaps.csv'

Both notebooks already guard for empty data (len(df) == 0 -> "No data available"), but the raw read crashes before reaching that guard.

Fix

• Switch the additionalDataTables reads to read_optional_csv() (the helper added in fix: handle missing DependenciesDeclared in release metro visualizations #104 for the release-metro notebooks), which returns an empty DataFrame for a None / missing / empty path. A missing optional table then flows into the existing "No data available" guard instead of crashing.

• complexity_vs_test_gaps_bubble: data_file_gaps; also drop the now-unused import pandas as pd.

• code_quality_executive_dashboard: data_file_methods, data_file_packages, data_file_smells, data_file_gaps (keeps pandas, still used by pd.concat / pd.Series).

Test plan

• papermill with a missing/nonexistent data_file_gaps -> runs to completion and renders the "No data available" figure (previously PapermillExecutionError / FileNotFoundError).
• papermill with real ClassQualityMetrics + TestGaps CSVs -> renders the bubble chart (2,137 points).
• ruff check clean on both changed notebooks.
• poe check-options passes for both.

Note

• This is resilience hardening, matching fix: handle missing DependenciesDeclared in release metro visualizations #104.