Lift tryfinally to macros to fix user code locations

[IanButterworth]

This fixes --track-allocations=user

        0             @timeit to "processing outputs" @views for out in yolo.out
        0                 outnr += 1
     1600                 w, h, a, bo, ba = out[:size]
      480                 weights = reshape(yolo.W[out[:idx]], w, h, a, bo, ba)
        0                 sxy = out[:scale_x_y]
        -                 # adjust the predicted box coordinates into pixel values
        0                 if yolo.cfg[:output][outnr][:new_coords] == 1
        0                     if haskey(yolo.cfg[:output][outnr], :max_delta)
        0                         delta = Float32(yolo.cfg[:output][outnr][:max_delta])
        0                         clamp!(weights[:, :, 1:2, :, :], -delta, delta)
        -                     end
        0                     weights[:, :, 1:2, :, :] = (weights[:, :, 1:2, :, :] .* sxy .- (sxy - 1)/2 .+ out[:offset]) .* out[:scale]
        0                     weights[:, :, 3:4, :, :] = (weights[:, :, 3:4, :, :] .* sxy).^2 .* out[:anchor]
        -                 else
        -                     # Classic behavior
     6560                     weights[:, :, 1:2, :, :] = (σ.(weights[:, :, 1:2, :, :]) .* sxy .- (sxy - 1)/2 .+ out[:offset]) .* out[:scale]
     3040                     weights[:, :, 3:4, :, :] = exp.(weights[:, :, 3:4, :, :]) .* out[:anchor]

Before

    42768             @timeit to "processing outputs" @views for out in yolo.out
        -                 outnr += 1
        -                 w, h, a, bo, ba = out[:size]
        -                 weights = reshape(yolo.W[out[:idx]], w, h, a, bo, ba)
        -                 sxy = out[:scale_x_y]
        -                 # adjust the predicted box coordinates into pixel values
        -                 if yolo.cfg[:output][outnr][:new_coords] == 1
        -                     if haskey(yolo.cfg[:output][outnr], :max_delta)
        -                         delta = Float32(yolo.cfg[:output][outnr][:max_delta])
        -                         clamp!(weights[:, :, 1:2, :, :], -delta, delta)
        -                     end
        -                     weights[:, :, 1:2, :, :] = (weights[:, :, 1:2, :, :] .* sxy .- (sxy - 1)/2 .+ out[:offset]) .* out[:scale]
        -                     weights[:, :, 3:4, :, :] = (weights[:, :, 3:4, :, :] .* sxy).^2 .* out[:anchor]
        -                 else

[IanButterworth]

Needs more work for @timeit on function defs, which I've left as-is for now. That needs more thought.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

Attention: Patch coverage is 88.88889% with 4 lines in your changes missing coverage. Please review.

Project coverage is 94.57%. Comparing base (ada4b6c) to head (278e796).
Report is 15 commits behind head on master.

Files with missing lines	Patch %	Lines
src/TimerOutput.jl	88.88%	4 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #196      +/-   ##
==========================================
+ Coverage   84.82%   94.57%   +9.74%     
==========================================
  Files           5        6       +1     
  Lines         323      479     +156     
==========================================
+ Hits          274      453     +179     
+ Misses         49       26      -23     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[IanButterworth]

@ericphanson do you know whether this ontop of #194 is a full fix, or should we try to take this approach for function defs too?

[ericphanson]

I'm not sure, I don't fully understand either fix aha. I think #194 is basically removing the wrong line nums and pushing in the source one. For this one, what does tryfinally have to do with it?

[IanButterworth]

The issue this fixes is that doing the Expr(:tryfinally deep in the functions called by the macros was overwriting the source info for the user code. Doing the Expr(:tryfinally at the top level macro avoids that issue.

@KristofferC any opinion here? I'm proposing merging this and then rebasing #194 and merging if green, given that PR has tests.