Fix URL encoding for non-ASCII bytes

[fallintoplace]

Summary

• Treat URL encoder input bytes as unsigned before splitting them into hex nibbles.
• Add UTF-8 regression coverage for café encoding to caf%C3%A9.

Why

On platforms where char is signed, bytes >= 0x80 can become negative before the hex lookup. That can lead to undefined behavior or incorrect percent encoding for non-ASCII strings.

Testing

• cmake -S . -B build -G Ninja -DICEBERG_BUILD_BUNDLE=OFF -DICEBERG_BUILD_REST=OFF -DICEBERG_BUILD_HIVE=OFF -DICEBERG_BUILD_SQL_CATALOG=OFF
• cmake --build build --target util_test
• ctest --test-dir build -R util_test --output-on-failure

[zhjwpku]

Good catch, for non-ASCII input bytes, kHexChars[c >> 4] can lead to out-of-bounds access and undefined behavior. +1 for this fix.