Avoid issues when converting large integers to floating point (bug #62212). * libinterp/corefcn/oct-stream.cc (get_size), libinterp/corefcn/xpow.cc (xisint), libinterp/octave-value/ov-base.cc (INT_CONV_METHOD), libinterp/octave-value/ov.cc (check_colon_operand, range_numel (T, double, T)), liboctave/numeric/lo-mappers.cc (nint_big, nint), liboctave/util/oct-inttypes.cc (emulate_mop, operator - (const double, const octave_uint64)), liboctave/util/oct-string.cc (rational_approx): Take into account that the maximum value of (signed or unsigned) integers might change its value if converted to floating point. In comparisons, check against the first value *outside* the range of the integer type instead of the last value *inside* its range.