Symbolica

**Notable changes**
- User-defined normalization functions (see [here](https://reform.zulipchat.com/#narrow/stream/392700-functionality/topic/normalization/near/469959172))
- Arbitrary depth dual numbers (see [here](https://reform.zulipchat.com/#narrow/stream/392700-functionality/topic/dual.20numbers/near/472259682))
- Non-linear system solving using Newton's method (see [here](https://reform.zulipchat.com/#narrow/stream/392700-functionality/topic/numerical.20solving/near/474013970))
- `symb!()` and `fun!()` improvements for faster symbol and function construction (see [here](https://reform.zulipchat.com/#narrow/stream/363562-api/topic/rust/near/472431296))
- A vector type (Rust only for now)

**Notable changes**
- Graph generation
- Graph API for Pyton
- Native Windows support! You can now `pip install symbolica` on native Windows.
- Introduce short-hand notation `S`, `E`, `N` to define symbols, expressions and numbers in Python
- Fused `symbols` and `symbol`
- Right-hand side of replacement can now be a function
- Pattern filter based on the matches so far
- Determine size of automorphism group of graphs
- Support antisymmetric tensor canonization

**Upgrade notice**
- `Expression.symbols` has been deprecated, you can now always use `Expression.symbol`. You can also just write `S`.
- The RHS of replacements is no longer a pattern and requires an extra `into()` to convert from a pattern
- The match conditions also require an extra `into()` to support match stack filter functions

**Notable changes**
- Tensor dummy index symmetrization
- Graph canonization, with support for directed edges and vertex and edge colours
- Add support for cyclesymmetric functions
- ASM support for x86_64 Mac

**Notable changes**
- Optimized nested expression evaluation with C++ and ASM output modes
- Seamlessly load compiled evaluation library
- Add cycle-symmetrization transformer
- Improve Rust ergonomics

**Notable changes**
- Support for floating point coefficients, parsing of floats in standard and exp-notation
- Precision tracking floating point numbers
- Galois fields (also in Python API)
- Factorization of polynomials over Galois fields and algebraic numbers
- Extension of algebraic numbers, e.g. joining `sqrt(2)` and `sqrt(3)`
- Linearization transformer

**License changes**
- Non-commercial users are now allowed to run Symbolica on one core and instance per device for **free** without requiring a sign-up
- License keys will now work offline. Remember to [update the key](https://symbolica.io/docs/get_started.html#updating-the-key)!

**Notable changes**
- Multiprecision floats can be used for evaluations
- Introduce error propagating float
- Series can now be constructed using a relative depth

**Upgrade notice**
- Offline keys have been deprecated
- `rug::Float` does not implement `Real` anymore. Use `.into()` to convert to the wrapper type `Float` that does operations that do not lose precision (see [here](https://gitlab.com/tspiteri/rug/-/issues/64))