Refactor timing routines

[conradtchan]

Type of PR:
modification to existing code

Description:
Makes the timing hierarchy clearer using a tree diagram. E.g.

                    wall        cpu  cpu/wall     frac
 ├─step        :   1.69s      7.15s     4.24   100.00%
 │ ├─tree      :   0.28s      0.39s     1.40    16.67%
 │ │ └─balance :   0.16s      0.04s     0.26     9.26%
 │ ├─density   :   0.66s      1.18s     1.81    38.89%
 │ ├─force     :   0.59s      1.54s     2.60    35.19%
 │ └─extf      :   0.12s      0.24s     1.91     7.41%
 ├─write_ev    :
 └─write_dump  :

This change makes it easier to perform and visualise detailed timing of communication routines.

• A timer for MPI balancing in tree building has been added to provide an example of multiple nested timers
• Tree diagram is generated automatically using the hierarchy defined in setup_timers
• All timers are stored in an array of timers in one place, rather than scattered across the code

Testing:
Examined printout from running a simple disc problem with MPI

Did you run the bots? yes

[danieljprice]

all looks good from my side, feel free to merge :)