[ptensor] add missing elementwise binary operations

The following elementwise binary operations need an implementation in PTensorToLinalg(.cpp) (see [array-API spec](https://data-apis.org/array-api/latest/API_specification/elementwise_functions.html) for the expected behavior of the operations):
* [ ] ptensor::ATAN2
* [ ] ptensor::LOGADDEXP
* [ ] ptensor::LSHIFT
* [ ] ptensor::MATMUL
* [ ] ptensor::TRUE_DIVIDE
* [ ] ptensor::BITWISE_AND
* [ ] ptensor::BITWISE_LEFT_SHIFT
* [ ] ptensor::BITWISE_OR
* [ ] ptensor::BITWISE_RIGHT_SHIFT
* [ ] ptensor::BITWISE_XOR
* [ ] ptensor::EQUAL
* [ ] ptensor::GREATER
* [ ] ptensor::GREATER_EQUAL
* [ ] ptensor::LESS
* [ ] ptensor::LESS_EQUAL
* [ ] ptensor::LOGICAL_AND
* [ ] ptensor::LOGICAL_OR
* [ ] ptensor::LOGICAL_XOR
* [ ] ptensor::NOT_EQUAL

See also [Operation Details->Elementwise Operations in the RFC](https://github.com/intel/mlir-extensions/tree/refactor/docs/rfcs/20220804-ptensor#operation-details).

Reference implementations can be found in the TOSA dialect and/or on main (numba_dpcomp/numba_dpcomp/mlir/numpy/funcs.py [Python]).

It is ok to let initial implementations operate on default PTensorTypes only, e.g. ignore device and distribution attributes of input tensors.