jobtap: add allocated resource information in job.state.run callbacks

[grondo]

We have two use cases where an external jobtap plugin needs access to the allocate resource information in R in the job.state.run or later callback.

1. The cray-pals plugin needs to determine if the number of nodes allocated to a job are > 1 in order to determine if ports need to be allocated to the job.
2. The flux-accounting fairshare plugin (possibly) needs access to the count of nodes allocated to jobs to enforce max_jobs limits

Currently, each of these plugins will need to (separately) fetch R from the KVS and process necessary information.

It may potentially help these and future use cases if R were returned directly from the scheduler to the job-manager in the alloc response, and made available (perhaps with some convenience functions) to jobtap plugins in the job.state.run callback.

I'm not sure if it would be better for the kvs_put() of R to also be moved to the job-manager, or if the alloc protocol should just send a copy of R or a subset of information.

At this point, this issue is just a starting point for a discussion.

[cmoussa1]

This is the issue I was trying to remember during our team meeting today - I believe this would be a prerequisite for flux-accounting to be able to enforce a max-nodes limit across all of their running jobs.

[grondo]

It appears this was actually fixed by @garlick in commit 4158ae2.