Create performance counters for primitives

[parsa]

This should be done in a way that all primitives almost automatically have performance counters associated with each of them. We should at least create two performance counters for each of the primitives:

/phylanx{locality#N/<primitive-name>}/invocation-count
/phylanx{locality#N/<primitive-name>}/execution-time

[hkaiser]

The names for the counters should follow the established practice. Here is a list of possible names for the counters we should create for each of the primitives:

/phylanx{locality#N/total}/<primitive>/count/eval
/phylanx{locality#N/total}/<primitive>/time/eval
/phylanx{locality#N/total}/<primitive>/count/direct-eval
/phylanx{locality#N/total}/<primitive>/time/direct-eval

where <primitive> represents the primitive type the counter will return data for.

All returned values should be cumulative, i.e. the values should represent the accumulated counts/times since the counter was reset last. The returned times should have nanosecond resolution.

Also, while discussing this, we decided that the counters should return a std::vector<std::int64_t> of values (i.e. the counter types should be hpx::performance_counter::counter_raw_values - note that this type does not exist yet, see STEllAR-GROUP/hpx#3102).

The values in the returned array should be ordered based on the sequence number that is encoded in the primitive names (as they are registered in AGAS), see #143 for a description of the structure of the generated primitive names.

Examples for the counter names used would be:

 /phylanx{locality#0/total}/define-function/count/eval
 /phylanx{locality#0/total}/add/time/direct-eval

[hkaiser]

@parsa please have a look at STEllAR-GROUP/hpx#3103, those changes will most probably be needed for your work. Please comment over there if that's what you need for your work on the performance counters.

[hkaiser]

This will also require for STEllAR-GROUP/hpx#3118 to be merged in order to be able to (re-)initialize the counters with the current list of instantiated primitives.

[hkaiser]

This was fixed by merging #165, #169, and #176