Improve CasToComparableText

[reckart]

Is your feature request related to a problem? Please describe.
When inspecting or diffing CAS contents for tests we frequently rely on a simple CSV stringification that:

• does not preserve rich, human-friendly output (HTML) for easier visual inspection,
• lacks configurable columns (anchor, covered text, indexed status),
• produces unstable ordering for multi-valued/annotation features and ambiguous anchors,
• offers no convenient way to exclude noisy features/types or treat empty strings specially,
• and forces long covered text into the output making diffs noisy.

I'm often frustrated when test failures produce long, hard‑to‑scan CAS dumps or when small, irrelevant differences (e.g., non-deterministic anchor numbering or list order) make comparisons brittle.

Describe the solution you'd like
Add an enhanced CAS -> comparable text utility with the following capabilities:

• Output formats: Keep CSV but add an HTML renderer for nicer human-readable tables.
• Configurable columns: enable/disable an anchor column, an indexed column, and a covered‑text column (with configurable max length and middle-abbreviation).
• Anchor formatting: anchors include type short name, optional annotation offsets, optional sofa id, optional indexing marker, and stable disambiguation suffixes for duplicate anchors; support optional anchor feature hash suffix.
• Stable ordering: when multi‑valued features hold annotations, optionally sort them by begin (asc), end (desc), type name to provide deterministic set‑like ordering.
• Index awareness: mark FSs as indexed and optionally add a dedicated <INDEXED> column; use indexed status as a tie-breaker when ordering.
• Exclusions: allow regex patterns to exclude specific features or types from rendering (cache regex compilation for performance).
• Null/empty handling: configurable nullValue, and an option to treat empty strings as null so empty values don’t clutter diffs.
• Multi‑valued rendering: robust handling of array/list FSs and primitive arrays, rendering them as bracketed lists; handle nested multi-valued structures recursively.
• Rendering options: omit XML declaration in HTML output and minimal inline styling so HTML is self-contained.
• Public API knobs: setters/getters for all above flags so callers can tune output for different use cases (compact machine diffs vs human inspection).

This produces a single stable, configurable comparable representation useful for both automated assertions and human debugging.