[bug] CrossSentenceMention needs start and end character offsets

[myedibleenso]

Character start and end offsets for CrossSentenceMentions can't be retrieved in a straightforward manner from the Odin Mention JSON we receive from clulab/processors:

clu-processors/python/lum/clu/odin/serialization.py

Line 103 in 11970d5

start = mjson["characterStartOffset"]

This is because a cross sentence mention involved multiple token intervals.

Proposed solution

1. Safely retrieve start and end via mjson.get("characterStartOffset", None) and mjson.get("characterEndOffset", None)
2. Construct CrossSentenceMention's

   clu-processors/python/lum/clu/odin/serialization.py

   Lines 215 to 216 in 11970d5

   	start=start,
   	end=end,

   using min and max token intervals for anchor and neighbor
3. Add unit tests (see this example as a reference)