Various bugfixes in the grammar

[apozharski]

Hello,
There is currently a bug in grammar (more accurately the scanner/lexer) which causes errors when a multivariable assignment contains a line continuation between the lhs and the equals sign. For example:

function [] ...
    = f_ellipsis_empty_output(arg)
end

this kind of function declaration would trigger this error.

This patch fixes this issue and changes the way that the multivariable assignment is parsed to remove the spurious ignored variable token when the lhs is empty.

[apozharski]

I have added additional changes to correctly handle trailing semicolons on various blocks. it may make sense to do this in a more elegant way via a shared "end" symbol which has an optional trailing semicolon. I will collate the remaining bugfixes arising from the test suite in sphinxcontrib-matlabdomain in this PR.

[acristoffers]

Thank you for your pull request, but I'll have to ask you to split it in two: one with the changes for the multi-output assignment of the original PR and another with the other changes pertaining the semicolons. The PR for the multi-output assignment is going to be accepted almost immediately, but the other one I want to test and take a better look at the changes to make sure there is no regression, since it will touch quite a few rules. And sorry for taking so long to answer, I was rather busy this week :(

[apozharski]

Yep, that is no problem, frankly they were only lumped together because of laziness. Closing this and will open separate PRs.