Fix signed integer overflow UB in BSNumber and BitHacks for INT_MIN

[CodeReclaimers]

BSNumber(int32_t) and BSNumber(int64_t) negate negative inputs via number = -number, which is undefined behavior for INT32_MIN/INT64_MIN. Fix: compute the magnitude in unsigned arithmetic.

BitHacks::GetTrailingBit(int32_t) has the same issue with value & -value for INT32_MIN, and the uint32_t overload delegated to it. Fix: reverse the delegation so uint32_t does the work using ~value + 1u, and int32_t delegates to uint32_t.

Test: build with -fsanitize=undefined and construct BSNumber from INT32_MIN and INT64_MIN.

Test (CMakeLists.txt)

cmake_minimum_required(VERSION 3.14)
project(test_issue_8_2 LANGUAGES CXX)
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED ON)
set(GTE_ROOT "${CMAKE_CURRENT_SOURCE_DIR}/../../GTE")
add_executable(test_issue_8_2 test_issue_8_2.cpp)
target_include_directories(test_issue_8_2 PRIVATE "${GTE_ROOT}")

Test (test_issue_8_2.cpp)

#include <cstdlib>
#include <cstdint>
#include <climits>
#include <cmath>
#include <Mathematics/BSNumber.h>
#include <Mathematics/UIntegerAP32.h>

using BSN = gte::BSNumber<gte::UIntegerAP32>;

int main()
{
    BSN bn32(INT32_MIN);
    BSN bn64(INT64_MIN);
    double v32 = static_cast<double>(bn32);
    double v64 = static_cast<double>(bn64);
    bool ok = (v32 == static_cast<double>(INT32_MIN))
           && (v64 == static_cast<double>(INT64_MIN));
    return ok ? EXIT_SUCCESS : EXIT_FAILURE;
}