FastFloat's parsing of scientific notation strings for integers with specific digit counts can lead to precision loss

[forezp]

hi, FastFloat's parsing of scientific notation strings for integers with specific digit counts can lead to precision loss. For example, when parsing '2.0250405e+07', the resulting value is 2.0250404999999996e+07.

Below are test cases：

func TestParse(t *testing.T) {
	s := "2.0250405e+07" //20250405
	t.Log("raw:", s)
	parsed, err := Parse(s)
	if err != nil {
		t.Fatal(err)
	}
	t.Log("Parse:", parsed)
	parsed = ParseBestEffort(s)
	t.Log("ParseBestEffort:", parsed)
	parsed, err = strconv.ParseFloat(s, 64)
	if err != nil {
		t.Fatal(err)
	}
	t.Log("strconv.ParseFloat", parsed)
}

output:

=== RUN   TestParse
    parse_test.go:12: raw: 2.0250405e+07
    parse_test.go:17: Parse: 2.0250404999999996e+07
    parse_test.go:19: ParseBestEffort: 2.0250404999999996e+07
    parse_test.go:24: strconv.ParseFloat 2.0250405e+07
--- PASS: TestParse (0.00s)
PASS