Fix mixed signed/unsigned comparisons and enforce size_t for object sizes in infer_shape

[lyriccoder]

In infer_shape, variables holding tensor element counts were declared as int/uint32_t and compared against fields from generic_param. This caused:

1. Mixed signed/unsigned comparisons → implicit conversions may lead to incorrect results.
2. Use of int/uint32_t for object sizes instead of the recommended size_t.

Importance

• May cause logical errors when values exceed INT_MAX.
• Potential risk of undefined behavior with very large tensors (>2G elements).

Related Standards:

• MISRA C++ Rule 5-0-4: "An implicit integral conversion shall not change the signedness"
• CERT C++ INT31-C: "Ensure that integer conversions do not result in lost or misinterpreted data"

Possible CWE Classification:
CWE-681: Incorrect Conversion between Numeric Types