Add option to ignore function coverage

[Spacetown]

If coverage isn't ignored, the JSON output is extended with property 'gcovr/function' with value True/False.

close #398

[codecov]

Codecov Report

Merging #430 (9aa0d49) into master (2aef434) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #430      +/-   ##
==========================================
+ Coverage   95.52%   95.54%   +0.01%     
==========================================
  Files          20       20              
  Lines        2459     2469      +10     
  Branches      420      423       +3     
==========================================
+ Hits         2349     2359      +10     
  Misses         48       48              
  Partials       62       62              

Flag	Coverage Δ
ubuntu-18.04	95.05% <100.00%> (+0.02%)	⬆️
windows-2019	95.13% <95.45%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
gcovr/configuration.py	99.28% <ø> (ø)
gcovr/tests/test_gcov_parser.py	100.00% <ø> (ø)
gcovr/gcov.py	88.13% <100.00%> (+0.34%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2aef434...9aa0d49. Read the comment docs.

[latk]

Missing a changelog entry, otherwise ready to go :) Thanks for working on this.

Potential future directions:

• The gcov parser state machine is growing more and more complex. Perhaps it would be possible to separate this into a textual part that generates events for each line, and then various event consumers with simpler state machines that update the coverage data structure. For example when reviewing this code, I was unsure where the last_was_function_marker was being reset because it was spread over so many lines.

• For each compiler version, we are generating all output formats. This requires extra effort when updating reference files and slows down the test suite. It would be sufficient to only test JSON output for individual compilers, since the other outputs can be generated from JSON without depending on any compiler. So we could move from O(compilers × outputs) tests to O(compilers + outputs). In some places, the test suite has already been rewritten to use JSON as an intermediate format but the JSON isn't reused between tests and the tests still depend on freshly generated JSON (requiring a compiler).

[Spacetown]

@latk I've rebased and added the issue to the changelog.
The event based parser is a good idea but first I'll create an abstract interface for readers and writers as discussed in #443.
I think we can't remove the call of gcc in the whole test suite because also the gcov parser must be tested. But the tests of the writer can be improved here.

[latk]

Building the sphinx docs found an error for us:

Suggested change

	- Add :option:`--exclude-function-coverage` to exclude the line of the function definition in the coverage report. (:issue:`430`)
	- Add :option:`--exclude-function-lines` to exclude the line of the function definition in the coverage report. (:issue:`430`)

[Spacetown]

Yes, currently I'm rebasing and updating the test data.