float_to_integer_helper doesn't handle fractions correctly

[kevinbackhouse]

Found by OSS-Fuzz: https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=45291

OSS-Fuzz has reported this as an integer overflow in sumToLong. The actual problem is in float_to_integer_helper:

exiv2/include/exiv2/value.hpp

Lines 1248 to 1258 in 5d5354e

	//! Utility for toInt64, toUint32, etc.
	template <typename I>
	inline I float_to_integer_helper(size_t n) const {
	const auto v = value_.at(n);
	if (static_cast<decltype(v)>(std::numeric_limits<I>::min()) <= v &&
	v <= static_cast<decltype(v)>(std::numeric_limits<I>::max())) {
	return static_cast<I>(v);
	} else {
	return 0;
	}
	}

It seems that the static_cast<I>(v) produces garbage when v is a fraction. In the poc, v == 6.0134700169990685e-154.