Add RISC-V target detection

[carlosqwqqwq]

Why

base64 did not recognize RISC-V explicitly in its CMake target detection, so riscv configurations could depend on generic fallback behavior and unclear SIMD exclusion. This patch makes the baseline RISC-V path explicit while keeping the existing portable codec path.

What changed

• Detect RISC-V targets in the CMake architecture probe through __riscv and __riscv_xlen.
• Recognize riscv64, rv64, riscv32, rv32, and generic riscv* values from CMAKE_SYSTEM_PROCESSOR.
• Generate BASE64_TARGET_RISCV for RISC-V CMake configurations.
• Keep RISC-V builds on the existing portable codec path, with SSE, AVX, AVX-512, NEON32, and NEON64 disabled.

Verification

• Ran native CMake configuration with tests and CLI enabled: cmake -S . -B build-codex-native -DBASE64_BUILD_TESTS=ON -DBASE64_BUILD_CLI=ON -DBASE64_WERROR=OFF.
• Built the native project successfully with cmake --build build-codex-native --parallel 2.
• Ran ctest --test-dir build-codex-native --output-on-failure; both test_base64 and benchmark passed.
• Ran simulated RISC-V CMake configuration with CMAKE_SYSTEM_NAME=Linux, CMAKE_SYSTEM_PROCESSOR=riscv64, and CMAKE_TRY_COMPILE_TARGET_TYPE=STATIC_LIBRARY.
• Built the simulated RISC-V portable static library successfully with cmake --build build-codex-riscv64-4 --parallel 2.
• Checked generated config.h: BASE64_TARGET_RISCV=1; BASE64_WITH_SSSE3, BASE64_WITH_SSE41, BASE64_WITH_SSE42, BASE64_WITH_AVX, BASE64_WITH_AVX2, BASE64_WITH_AVX512, BASE64_WITH_NEON32, and BASE64_WITH_NEON64 were all 0.
• Checked compile_commands.json: no -mavx, -mssse3, -msse4, -mfpu=neon, -march=arm, or /arch:AVX flags were present.

Notes

This is a conservative CMake portability patch. It does not add RVV codecs or RISC-V-specific Base64 SIMD acceleration. A real riscv64 cross-build or hardware run was not performed on the local Windows host because no RISC-V toolchain is available here.