Julia v0.7/1.0 support and drop v0.6 support

[iblislin]

TODO:

• new broadcast mechanism

• fix @nd_as_jl

• Add CI build for Julia 0.7

• Add CI build for Julia 1.0

• maybe another PR for docs.

[Roshrini]

@iblis17 Thanks for this contribution.
@mxnet-label-bot [pr-work-in-progress]

[ankkhedia]

@iblis17 Could you please update the PR description and check into the CI failure and change the title when the PR is good to be reviewed.

[iblislin]

yeah, will do that. but this PR isn't ready.

[stu1130]

@iblis17 Thanks for the amazing contribution
if you need any help, feel free to ask.
Is PR still working in progress?

[iblislin]

yeah, still WIP. (i'm just quite busy this week.)

[vandanavk]

@mxnet-label-bot add [pr-work-in-progress]

[iblislin]

Ready for review.

[iblislin]

@mxnet-label-bot update [julia, pr-awaiting-review]

[iblislin]

I'm going to merge it this weekend if no objection.

[marcoabreu]

LGTM, thanks a lot!

One minor thing where I'd like to get Aarons opinion one, but from CI side is everything good to go.

[iblislin]

I just found that the main doc build is deployed to an aws site.
Maybe Julia doc can be deployed to that site as well?

[iblislin]

The CI failures are unrelated.

[aaronmarkham]

@iblis17 Hi there - I was just tracing where the julia docs end up... and it seems that they are made but are not deployed anywhere yet... or am I missing a step that picks up outside of the CI process?

[iblislin]

@aaronmarkham I plan to deploy it in downside channel -- gh-page of MXNet.jl (https://github.com/dmlc/MXNet.jl)

Since the syntax needed by Julia is not support by the implementation of sphinx markdown plugin,
it's hard to integrate with current MXNet doc system.

[aaronmarkham]

@iblis17 Can you point me to more info about this Sphinx markdown plugin issue?

What about getting Julia into the new pipeline and website proposed here?
#14330

Could it follow the pattern that R using now? Or do we still have some plugin issue to deal with?

[iblislin]

@aaronmarkham well, I have tried to integrate it ... several month ago, but ran into markdown parsing failure.
I can set a experimental branch up, and try to re-produce the failure.