Stop using uint64 types

[stuarteberg]

We should start using int64 instead of uint64 wherever possible, since implicit conversions between the two can trip up commonly used libraries, and also cause subtle problems when serializing to language-agnostic formats, such as json, protobuf, and sqlite.

More crucially, the use of uint64 prevents us from using the numexpr library (or even having it installed), which is the preferred engine for pandas.DataFrame.query().

In practice, we never use the sign bit in our sv/body IDs, so there should be no problem switching to int64.