Improve Support for Arrays and Array-Like Expressions

Improve Type Inference for Array Literals

The Typifier currently has only limited support for inferring the type of array initializer lists (PsArrayInitList) when no type information about their entries is given. This is especially the case for declarations of arrays of numeric constants, e.g.

ps.Assignment(sp.Symbol("arr"), (1, 2, 3, 4))

should translate to C code as

double arr[] = { 1., 2., 3., 4. };

provided that float64 is the default_dtype. Currently, the type of arr cannot be inferred. When array literals such as this occur inline within an expression, such as x + {1, 2, 3, 4}[idx], type inference is also unreliable.

Requirement: If the type of an array literal's entries, and therefore the array itself, cannot be inferred from its entries, it should be inferred from the enclosing context in the same way as for scalar constants.

Parsing of SymPy Matrices and Tensor-Like objects

Currently, pystencils only parses sp.Tuple as an array literal. The implementation of FreezeExpressions should be extended such that immutable SymPy matrices and tensor-like objects (which qualify as expressions) are also parsed as array literals.

Documentation of Array-Parsing Capabilities

The new capabilities of pystencils to handle array-like objects have to be documented. Add a section about arrays to the symbolic language documentation.