Add ignoreCase and ignoreWhitespace options

[krkarma777]

Closes #16

Summary

• `ignoreCase`: tokens compare case-insensitively (all modes, including char via the generic token pipeline)
• `ignoreWhitespace`: whitespace runs compare equal; `line` mode compares trimmed lines (diffTrimmedLines-style); whitespace with no counterpart still diffs as insert/delete
• Both implemented as an interning-layer normalizer — the Myers core is untouched; the fast scanner paths still serve the default (no-options) case
• Semantics documented: with an ignore option active, `equal` texts come from `b`, so non-delete concatenation reproduces `b` exactly (a single equal text cannot be both `Fox` and `fox`)
• `diffTokens` accepts both options for custom tokens; `refine` sub-diffs inherit them

Test Plan

• 8 new tests incl. 1,800-case fuzz asserting the b-side reconstruction invariant across word/char/line x option combinations
• One test expectation was corrected during development (word-mode tokens don't split `ab`), caught by running the suite — the library behavior was right
• Full suite 58/58, typecheck clean