fearless_simd v0.7: 64-bit integers, improved generics, SSE2, and upcoming v1.0

Fearless SIMD version 0.7 introduces 64-bit integer support, an explicit SSE2 level, improved generic programming capabilities, and additional implemented SIMD operations. This marks the final major release before the upcoming v1.0. The API is now fully exposed for u64 and i64 vector types, which were previously unsupported due to hardware limitations.

Earlier versions did not support 64-bit integer vectors as many operations on them lacked hardware support, requiring emulation. However, in v0.5, the compiler now keeps track of which intrinsics are part of which instruction set, simplifying implementation and reducing unsafe blocks. The improvements in generic programming make writing functions generic over the vector type much more pleasant, allowing more algorithms to be expressed generically.

The SSE2 level now has operations expressed in terms of SIMD intrinsics, rather than relying on autovectorization on systems without SSE4.2. Despite the added features, compilation time remained unchanged, thanks to build profiling and optimization efforts. The type-specific reinterpret_* methods were removed in favor of the generic bitcast method, and the load_interleaved_128 operation was renamed for clarity.

With over a dozen direct dependents on crates.io and a thousand repositories using it, the design of the crate has been proven in real-world projects. If no API concerns arise, Fearless SIMD v1.0 is set to be released in early September, with no further changes planned for the v1.0 release.