- Update to version 0.6.4+0:
* Safety critical bounds checking is off-by-one in sgrproj_box_ab_internal
* Initialize `low` array in a more rust-like way in `kmeans`
* Rework mutable borrows for symbol_with_update
* Drop explicit size for macro symbol_with_update
* Use const generics for CDFContextLog
* Fix undefined behavior in CDFContextLogOps
* ec: Simplify lr_compute function
* Use a bit counter instead of a byte counter in WriterRecorder
* Minimize bounds checks in pred functions
* Use generics for BD-8 on sgrproj functions
* Use saturating_sub in sgrproj_sum_finish
* Move bounds checks out of hot loop in sgrproj
* Ensure quantizer values are non-zero with a const fn
* Optimize base quants using NonZero integers (#3115)
* Move quant tables to separate file to improve organization (#3113)
* Use is_power_of_two method in divu_gen
* Fix rounding issue in HBD CDEF code
* Hint that creating a region from an empty plane is unlikely
* Minor optimization to take_slice
* Clean up cdef_dist
* Minor optimizations
* Optimizations for weighted_sse
* Add HBD AVX2 assembly for SAD (#3099)
* ec: Manually inline `msb()` and `ilog()` for clarity (#3104)
* Comment regarding cnt being unused
* Improve naming
* Use a bit counter instead of a byte counter in EC
* Add 10-bit cdef_dist ASM
* Optimise sad_plane_internal
* Save some bounds checks on me_stats
* Minor optimization to av1_get_mv_joint
* Use chunks_exact for performance in diff method
* Integrate CfL AC x86 assembly functions
* benches: Fix alignment issue for cfl_rdo
* Move luma_ac to predict module
* Extract luma_ac_internal with const generics
* Optimise plane::as_region
* Optimise BlockSize::from_width_and_height_opt
* Improve vectorization in get_sad
* Template entropy coding functions to help optimiser
* Enable SSE2/AVX512ICL put/prep/avg x86 assembly
* Enable AVX2 12-bit Inverse Transform x86 assembly
* Enable new SSE4.1 HBD Inverse Transform x86 assembly
* Pin assert_cmd, predicates and clap in Cargo.toml for rust 1.60.0
* Prepare for release
* Fix header coding for level_idx < 7.