Remove dependency on ZipFiles.jl and EzXML.jl

[TimG1964]

This PR follows previous attempts and helpful feedback on Discourse here, and here. There was another thread, too, but I can't find it now. (EDIT Not a discourse thread but comments on a previous PR, here. I think I've followed all of @nhz2's advice in this PR.)

I've removed all dependence on both ZipFiles.jl (in favour of ZipArchives.jl) and EzXML.jl (in favour of XML.jl). I could not replace ZipFiles until after I had replaced EzXML, but EzXML was used ubiquitously throughout the XLSX package so replacing it was a comparatively big task.

Some important caveats:

• My skill levels are low
• I am unfamiliar with the internals of XLSX files
• I am unfamiliar with XML

I have relied on the heavy lifting done by @felipenoris throughout.

This PR passes all tests locally and works perfectly in my use case, which gives some confidence in making this PR. However, I think it will still need review by a competent person both for correctness and for efficiency. I haven't benchmarked, but it seems, subjectively, to be about the same speed as before.

[TimG1964]

There are a couple of other things to point out:

• I had a problem with the escaped characters (&<>"'). There was an element of coding to the test here to get the tests to pass but I also changed the tests a little. I had particular problems with trailing apostrophes! Although things now appear fine, and I think (mis) use of these characters (e.g. in tab names or column headers) is rare, some further validation might be valuable.
• I've changed the escape.xlsx test file. I think the sheet name in the original was illegal (too long), so I've shortened it to a valid length.
• The previous runtests.jl script itself relied on the EzXML.findall function, and there is no equivalent in XML.jl. I've included my own code in the runtests.jl script to address this but recognise that using my own code to check itself isn't ideal!