Add TOSA to TTIR conversion for ConstOp

[sdjukicTT]

tosa.const (mlir::tosa::ConstOp)

Can be lowered directly to ttir.const op.

Requires rebuilding of the value attribute in the following cases:

• Scalar values: TTNN does not support scalar types. So they should be converted to 1-D tensors.
• Boolean tensor: TTNN does not support boolean data. So they should be converted to bfloat16 tensors.
• Integer tensor: TTNN does not support 48 bit integer. So they should be converted to 32 bit tensor.
• Float tensor: TTNN does not support 64 bit float. So they should be converted to 32 bit tensor.

Methods from StableHLOToTTIRConstantOpConversionPattern in lib/Conversion/StableHLOToTTIR/StableHLOToTTIRPatterns.cpp can be reused and should be moved to a file from which they can be used for both TOSA and StableHLO.