Add if_tsoffset_ns option to IDB

[GioMusto]

This new option allows having a grater precision in the timestamp offset of packets.

[guyharris]

So, in effect, the rules are:

1. initialize a pair consisting of a 64-bit seconds value and a 32-bit nanoseconds value to {0, 0};
2. if there's a ts_offset option, set the seconds value to the option's value;
3. if there's a ts_nsoffset option, set the nanoseconds value to the option's value;
4. for all packet from this interface, add the resulting seconds/nanoseconsds value to each time stamp.

What should be done if the time stamp resolution for this interface isn't 1 nanosecond? For example, what if it only has microsecond resolution?

An alternative would be to have an offset option that's a 64-bit value, with the same resolution as interface time stamps on the interface; you'd just add that value to the time stamp in a packet block; that seems a bit less complicated. The ts_offset value, if any, would be an additional offset to add, after the 64-bit offset is added.

(Having ts_offset be in seconds would provide a way to keep the pcapng file format usable for microsecond-resolution time stamps even after 2^63-1 microseconds (about 292271 years), usable for nanosecond-resolution time stamps even after 2^63-1 nanoseconds (about 292 years) and, if we ever need picosecond-resolution time stamps, even after 2^63-1 picoseconds (about .29 years).)

[GioMusto]

I also thought about the second option (having a separate 64-bit value), but I didn't like it because it provided potential overlap between the two options (i.e. multiple ways of expressing the same offset), but I agree it would provide more flexibility.

For that to work, we should still define a default precision in case there's no if_tsresol option. Also, is it ok having an option depending on another one?