Improve handling of ris-live HTTP errors

[pforemski]

Hi,

While debugging ris-live errors (HTTP ERROR: Transferred a partial file (18)) I found a few small issues in the code, fixed/improved in the below 2 commits.

• fc71272 fixes an integer overflow error, where a timestamp in milliseconds was written to an uint32_t, which somehow didn't manifest for me
• 44fb56e is a poor-man's solution to no HTTP read error handling in ris-live (I was really surprised to see this guys... :P) see also Handle ris-live stream error messages #139

Thanks for your great work!

[alistairking]

Hey!
Thanks for taking the time to look into this, and for the PR.

The integer overflow bug is a nice catch. Thanks.

I'm not too sure about handling the read error like that, so I'll have to spend some quality time with the code to double check that it doesn't have unwanted side effects (it's probably ok, but since that code is right in the core of bgpstream, i just want to be extra careful).

I do think however, that when we get EOF from the transport (the 0 return), then we should return END_OF_DUMP. Remember that we don't know what transport we are reading the data over, so it is perfectly reasonable to be reading ris-live formatted data from a file (indeed this is how we test things).