snmp-ups: enhance the `--mibs` (plural!) to actually support several MIBs

[jimklimov]

With the argument called mibs (plural) it could make sense to actually support a list of MIBs to try to get info from, in order of preference, e.g. both a vendor MIB and the IETF MIB for missing data points. Or even several vendor MIBs for devices from companies with a long heritage.

After all, the running driver instance builds itself an in-memory list of OIDs to walk (optionally minding ways convert data) which pertain to the certain device. Why not pick lines originally stored in different mib2nut tables?..

@aquette : WDYT? Any blockers I've overlooked here? Theoretically this is something that DMF (#182, branch et al) could achieve with specially crafted mapping resource files, so why can't the core do similar with its built-in data?

Loosely related to issue #273 (about general driver multi-plexing).