Improve performance when handling large XML files with significant column augmentation

[ollietulloch]

The Xml::Table class attempts to augment column_mappings based on the data where there might be repeating sections/fields. It does this for each line, and so can be memory/compute intensive.

This PR takes a number of small steps to improve the efficiency, focusing on the Xml::MaskedMappings class.

Before this PR, ~1250 repeating sections took ~95 seconds for mapping augmentation. This PR gets that down to ~5.8 seconds.

Some of the small steps...

• Frozen object reuse e.g. DO_NOT_CAPTURE_MAPPING = { 'do_not_capture' => true }.freeze
• Using deep_dup only when needed
• Pre-allocating Arrays and hash with exact size to avoid resizing operations
• Direct array assignment to avoid intermediate objects (e.g. .map)
• Where possible, using O(1) lookups instead of O(n) searches
• Use Sets

[kenny-lee-1]

LGTM
Thanks for looking at it