printf: Parsing large numbers outside of type range should return saturated value wrapped in the error

[drinkcat]

When parsing very large numbers into i64, we should return an error, but also wrap INT64_MIN/MAX value in the error so that value gets printed:

$ env printf '%d\n' -92233720368547758080
printf: ‘-92233720368547758080’: Numerical result out of range
-9223372036854775808
$ env printf '%d\n' 92233720368547758080
printf: ‘92233720368547758080’: Numerical result out of range
9223372036854775807
$ ./coreutils-main printf '%d\n' -92233720368547758080
printf: '-92233720368547758080': Numerical result out of range
0
$ ./coreutils-main printf '%d\n' 92233720368547758080
printf: '92233720368547758080': Numerical result out of range
0

Similarly with u64, the value returned is UINT64_MAX:

$ env printf '%u\n' 92233720368547758080
printf: ‘92233720368547758080’: Numerical result out of range
18446744073709551615
$ ./coreutils-main printf '%u\n' 92233720368547758080
printf: '92233720368547758080': Numerical result out of range
0

And again with f64 (ExtendedBigDecimal can also hit this as it is limited to 64-bit exponent . -- Note that scientific format is not supported by printf yet, #7474), where underflow returns 0 and overflow returns inf:

$ env printf '%f\n' 1e-92233720368547758080
printf: ‘1e-92233720368547758080’: Numerical result out of range
0.000000
env printf '%f\n' 1e92233720368547758080
printf: ‘1e92233720368547758080’: Numerical result out of range
inf

This float case also matters for seq, which does this correctly already, but I'm trying to merge the parsing functions.

[drinkcat]

1e104b7 fixes these, and adds tests as well.