Add block arithmetic operations

[tran-go]

Purpose

Issue: #56

This PR adds block matrix arithmetic support on top of the block tensor classes and constructors.

It enables large matrices and vectors to be split into encoded blocks when the logical input requires more slots than one ciphertext can provide.

Scope

1. This PR adds block support for:

• + / add;
• - / subtract;
• * / elementwise multiply;
• matrix-vector multiplication, matrix @ vector;
• matrix-matrix multiplication, matrix @ matrix;
• negation;
• transpose;
• cumulative sum;
• sum.
• component-wise function execution

2. Rename cumulative_sum to cumsum for NumPy API compatibility.
3. Rename pow to power for NumPy API compatibility.
4. Move auxiliary arithmetic helpers to arithmetic_utils.py.
5. Move block arithmetic logic and helpers to block_arithmetic.py and block_arithmetic_utils.py.

Out of scope

This PR also does not add full NumPy slicing, broadcasting, or general block reshaping support.

Validation

Added and ran simple block arithmetic tests covering:

• block vector arithmetic;
• block matrix arithmetic;
• block matrix-vector multiplication;
• oversized matrix/vector inputs that require block packing.