Fix total deltas

[IanButterworth]

Pulled out of #184 by @BenCurran98

This PR

julia> begin
       to = TimerOutput()
       @timeit to "1" sleep(1)
       print_timer(to)
       sleep(1)
       print_timer(to)
       end

────────────────────────────────────────────────────────────────────
                           Time                    Allocations
                  ───────────────────────   ────────────────────────
Tot / % measured:      1.00s / 100.0%              384B / 100.0%

Section   ncalls     time    %tot     avg     alloc    %tot      avg
────────────────────────────────────────────────────────────────────
1              1    1.00s  100.0%   1.00s      384B  100.0%     384B
────────────────────────────────────────────────────────────────────
────────────────────────────────────────────────────────────────────
                           Time                    Allocations
                  ───────────────────────   ────────────────────────
Tot / % measured:      1.00s / 100.0%              384B / 100.0%

Section   ncalls     time    %tot     avg     alloc    %tot      avg
────────────────────────────────────────────────────────────────────
1              1    1.00s  100.0%   1.00s      384B  100.0%     384B
────────────────────────────────────────────────────────────────────

Master (note the total time is erroneously 2s in the 2nd print)

julia> begin
       to = TimerOutput()
       @timeit to "1" sleep(1)
       print_timer(to)
       sleep(1)
       print_timer(to)
       end
────────────────────────────────────────────────────────────────────
                           Time                    Allocations
                  ───────────────────────   ────────────────────────
Tot / % measured:      1.07s /  93.5%           4.50MiB /   2.3%

Section   ncalls     time    %tot     avg     alloc    %tot      avg
────────────────────────────────────────────────────────────────────
1              1    1.00s  100.0%   1.00s    105KiB  100.0%   105KiB
────────────────────────────────────────────────────────────────────
────────────────────────────────────────────────────────────────────
                           Time                    Allocations
                  ───────────────────────   ────────────────────────
Tot / % measured:      2.10s /  48.0%           5.43MiB /   1.9%

Section   ncalls     time    %tot     avg     alloc    %tot      avg
────────────────────────────────────────────────────────────────────
1              1    1.00s  100.0%   1.00s    105KiB  100.0%   105KiB
────────────────────────────────────────────────────────────────────

[KristofferC]

To me showing 2s is correct. It's the time and allocations from when you created the timer. We don't start measuring the time from the first section but from when the time is created. I feel the same applies here.

[KristofferC]

Ref #36

[IanButterworth]

Ok. I've blocked registration JuliaRegistries/General#125432

[IanButterworth]

I think there's a good argument for the totals to be representative of the contents of the report, particularly if to is being saved and printed later. But #36 does sound like a good way to unambiguate intention.

I just thought this was a clear bug, so I'll revert and leave that to be resolved properly.