perf: reduce string normalization overhead

[cmilesio]

What

• Uses strings.TrimSpace for Trim so ASCII inputs take the standard-library fast path.
• Reworks NormalizeSpace around canonical-input detection and single-allocation ASCII/Unicode builders.
• Preserves Unicode and invalid-UTF-8 behavior without retaining oversized source buffers.
• Adds comparative fluent/stdlib benchmarks and stronger normalization fuzz coverage.

Why

The fluent wrapper itself inlines without adding allocations. The measurable cost was inside whitespace normalization.

Case	Before	After
Trim ASCII	~12.8 ns	~2.9 ns
NormalizeSpace ASCII	~370 ns, 552 B, 7 allocs	~114 ns, 48 B, 1 alloc
NormalizeSpace Unicode	~370 ns, 552 B, 7 allocs	~200 ns, 48 B, 1 alloc
Already normalized	~354 ns, 552 B, 7 allocs	~27 ns, 0 allocs

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!