**IMPORTANT**: We have changed serialization of subbands. The current version of the serialization is 3. We support
de-serialization of version 2 and 3. However, the serialization of version 2 is not supported anymore and to parse
archives you have to update to version 0.6.0.
Added
* DRIFT-605: Clang-format to pre-commit and CI, [PR-48](https://github.com/panda-official/WaveletBuffer/pull/48)
* DRIFT-617: Conan build for Linux armv8, [PR-49](https://github.com/panda-official/WaveletBuffer/pull/49/)
* ISSUE-53: Force conan installation before sdist build, [PR-54](https://github.com/panda-official/WaveletBuffer/pull/54)
* DRIFT-637: Add matrix_compressor support, [PR-55](https://github.com/panda-official/WaveletBuffer/pull/55/files)
* Pass through float precision as compression level, [PR-59](https://github.com/panda-official/WaveletBuffer/pull/59)
Changed
* DRIFT-605: Fix `.clang-format` include categories order, reformat cpp files, fix v15.0.4, [PR-48](https://github.com/panda-official/WaveletBuffer/pull/48)
* DRIFT-638: Change g++ version in conan profile, [PR-52](https://github.com/panda-official/WaveletBuffer/pull/52)
* BUG-56: Change conan version to 1.56, [PR-57](https://github.com/panda-official/WaveletBuffer/pull/57)
Removed
* DRIFT-617: OpenBLAS, [PR-49](https://github.com/panda-official/WaveletBuffer/pull/49/)
Fixed
* DRIFT-622: Incorrect WaveletBuffer decompositions serialization with sf_compression=0, [PR-50](https://github.com/panda-official/WaveletBuffer/pull/50)
* BUG-56: CalculateMaxDecompositionSteps(), unit test for kNone, [PR-57](https://github.com/panda-official/WaveletBuffer/pull/57)