feat(table): add data cleaning utilities (drop_na, fill_na, drop_duplicates, convert_types)

[AhmadYasser1]

Summary

This PR adds commonly requested data cleaning utilities to datascience.Table, addressing #656.

Changes

• Table.drop_na: drop rows/columns with missing values (any/all)
• Table.fill_na: fill missing values with a scalar or a strategy (mean/median/mode)
• Table.drop_duplicates: remove duplicate rows with subset/keep options
• Table.convert_types: convert via mapping (per-column callable or type) or simple inference for numeric strings
• Tests in tests/test_cleaning.py
• CHANGELOG update and version bump to 0.19.0

Motivation

Many workflows require basic cleaning before analysis. Providing first-class APIs in Table avoids round-tripping through pandas and keeps parity with typical data wrangling tasks for students.

Usage Examples

# Drop rows with any missing values in columns 'a' and 'b'
t2 = t.drop_na(subset=['a', 'b'])

# Drop columns that contain any missing values
t2 = t.drop_na(axis='columns')

# Fill with a scalar
t2 = t.fill_na(value=0)

# Fill numeric columns by mean; categorical via mode
t2 = t.fill_na(strategy='mean')

# Remove duplicate rows, keeping first occurrence
t2 = t.drop_duplicates()
# Or based on a subset, dropping all duplicates
t2 = t.drop_duplicates(subset='id', keep='none')

# Convert types with mapping or infer numeric types
t2 = t.convert_types(mapping={'id': int, 'price': float})
t3 = t.convert_types(infer=True)

Notes

• Uses numpy and pandas.isnull consistent with existing code; no hard pandas dependency at call sites.
• Non-intrusive to existing APIs; all methods return new tables (no inplace).

Checklist

• Tests added
• CHANGELOG updated
• Version bumped (delete if maintainers prefer to bump during release)