using TimerOutputs for duration/allocation measurements

[chmerdon]

The solve function now relies on TimerOutputs and measures each operator separately, the printing of the statistics can be steered with the new argument timeroutputs (default is true).

[pjaap]

Nice, that was a fast adoption :)
Although the timer object is called to in the README of TimerOutputs, I suggest to rename it to timer or tmr.

[pjaap]

Looks beautiful, I like it

───────────────────────────────────────────────────────────────────────────────────────────
                                                  Time                    Allocations      
                                         ───────────────────────   ────────────────────────
            Tot / % measured:                 24.8s /  96.2%           2.16GiB /  98.0%    

Section                          ncalls     time    %tot     avg     alloc    %tot      avg
───────────────────────────────────────────────────────────────────────────────────────────
assembly                              2    21.0s   87.8%   10.5s   1.65GiB   78.2%   847MiB
  BilinearOperator (first)            1    12.9s   54.1%   12.9s   1.20GiB   56.6%  1.20GiB
  LinearOperator (first)              1    2.80s   11.7%   2.80s    251MiB   11.6%   251MiB
  HomogeneousData (first)             1    801ms    3.4%   801ms    138MiB    6.4%   138MiB
  CombineDofs (first)                 1    206μs    0.0%   206μs   17.2KiB    0.0%  17.2KiB
  BilinearOperator (penalties)        1    698μs    0.0%   698μs    145KiB    0.0%   145KiB
  LinearOperator (penalties)          1   85.3μs    0.0%  85.3μs   17.2KiB    0.0%  17.2KiB
  HomogeneousData (penalties)         1    171ms    0.7%   171ms   8.96MiB    0.4%  8.96MiB
  CombineDofs (penalties)             1    3.96s   16.6%   3.96s   61.9MiB    2.9%  61.9MiB
  residual vector                     1    285ms    1.2%   285ms   6.54MiB    0.3%  6.54MiB
linear solver                         2    2.91s   12.2%   1.46s    473MiB   21.8%   236MiB
  initialization                      1   36.0ms    0.2%  36.0ms   2.31MiB    0.1%  2.31MiB
  solve! call                         1    2.40s   10.1%   2.40s    443MiB   20.4%   443MiB
  update solution                     2    375ms    1.6%   188ms   25.4MiB    1.2%  12.7MiB
  linear residual computation         1   69.2ms    0.3%  69.2ms   1.71MiB    0.1%  1.71MiB

[pjaap]

I vote for merging. A bit of fine tuning should be done but let's wait for user feedback.