fix(crypto): exclude timing test from coverage builds

The Elligator2 timing consistency test is now properly excluded from coverage
builds using cfg_attr(coverage, ignore). Coverage instrumentation adds significant
overhead that invalidates timing measurements:

Observed in CI coverage build:
- Normal build: mean=36ms, max_deviation=18ms
- Coverage build: mean=148ms, max_deviation=156ms (4x slower, 8.5x variance)

The test remains active in normal builds (including CI Test jobs) to catch
timing regressions, but is automatically skipped when running under coverage
instrumentation where timing measurements are meaningless.

This follows best practices for timing-sensitive tests in instrumented environments.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: doublegate

crates/wraith-crypto/src/elligator.rs

@@ -449,7 +449,11 @@ mod tests {
     /// This test verifies that decoding takes similar time for different inputs,
     /// which is an indicator of constant-time behavior. Note that true constant-time
     /// verification requires specialized tools (dudect, ctgrind).
+    ///
+    /// Timing tests are inherently incompatible with coverage instrumentation,
+    /// so this test is disabled when running under coverage (cfg(coverage)).
     #[test]
+    #[cfg_attr(coverage, ignore = "Timing tests incompatible with coverage instrumentation")]
     fn test_decode_timing_consistency() {
         use std::time::Instant;