librados: fix unitialized timeout in wait_for_osdmap

[cbodley]

ceph::timespan is not zero-initialized. use std::optional for an explicit empty state

this is causing RGW=1 vstart.sh -n to sometimes fail in radosgw-admin commands with:

2018-10-23 15:13:26.107 7efe9f50e280 -1 librados: timed out waiting for first osdmap from monitors
couldn't init storage provider

[mogeb]

Ah, although I did test that explicitly in unit tests but it seems like I still missed it. Looks good, there's an alternative though that is used in other places in the source code where the timespan is initialized to 0: ceph::timespan duration{0}; .

[cbodley]

thanks, i like that better. updated

[tchaikov]

thanks!

[cbodley]

this passed the rgw suite in http://pulpito.ceph.com/cbodley-2018-10-24_15:36:40-rgw-wip-cbodley-testing-distro-basic-smithi/ (failures are unrelated). does this need to go through any other suites?

[tchaikov]

@cbodley no, i think it's good to merge.

[cbodley]

thanks @tchaikov @mogeb!