use load instead of convert

[ggrossetie]

What I Did

• Add tests on Import.Image.getCurrentImagesDir
• Change the signature of getCurrentImagesDir to make it easier to test
• Fix the regular expression that checks if the string starts with a protocol: /^(?:[a-z]+:)?\/\//i (remote)
• Use the load function in attributeReferenceProvider.ts, tableOfContentsProvider.ts and image-paste.ts since we only need an Asciidoctor.Document (instead of the parse/convert)

In a following pull request, I will probably remove a layer of indirection since we don't really need both AsciidocEngine and AsciidocParser.
In other words, AsciidocEngine has little to no value.

[danyill]

Add tests on Import.Image.getCurrentImagesDir

I'm so happy to see some tests! 😀

In a following pull request, I will probably remove a layer of indirection since we don't really need both AsciidocEngine and AsciidocParser. In other words, AsciidocEngine has little to no value.

One of the problems with the existing approach is sometimes we've been directly calling the convertUsingJavascript and there are times we've been doing it through the AsciidocEngine which is instantiated on startup.
They're not the same and there are undesirable side effects on error messages and other things so I think this would be an improvement.