WIP: Julia 0.7, remove globals, missing value, scale_offset

[meggart]

This is work in progress that I started some time ago and unfortunately is not based on the last updates (eg. #59 , #55 etc.) This basically happened when I was working on the code anyway to do the julia 0.7 update and includes the following new features:

• runs on julia 0.7 without deprecation warnings
• should be more thread-safe by removing the pre-allocated global arrays and using MArrays and SArrays where appropriate, there
• also does not keep track of opened files anymore in global state, the high-level functions (ncread, ncwrite etc) all close the file after the operation is finished. The NetCDF.open function gets a do syntax to safely access files with the intermediate and array-like approach (see unit tests). This should make accessing the files more stable
• this is probably a merge/review nightmare, but I splitted up NetCDF.jl in several files to number of LOC per file to a reasonable limit
• add the replace_missing option to ncread and NetCDF.open, so that the new Array{T,Missing} can be read and written to/from files, this is done without making a copy of the data
• respect the scale_factor and add_offset attributes and apply at least readvar in-place. The way to make this work without changing the data types is by letting the NetCDF library do the conversion

I think this would fix the main concerns that were raised in #39

As I said, when I started this I did not foresee so much activity by other community members (thanks to @jmgnve ). I think one could either apply the last PRs #54 #57 #59 on top of this one. The problem is that I will offline for a few weeks and can only continue here end of July.

The other possibility would be to simply merge #59 and also do what @visr siggested JuliaGeo/meta#4 run nanosoldier to have a 0.7-compatible version as fast as possible. Then I would have to apply the changes proposed above. Let me know what you think, and if I don't reply, fell free to merge the other pull requests as needed.

[visr]

Ha nice! This is indeed pretty extensive :)

How about we merge #59 to master first, tag another 0.6 release, then rebase this branch on top of it.

The other open PR, #60 is code formatting. This will be more work to rebase on, so better we just apply it again after this PR is in. It was done automatically right, @jmgnve? Which tool did you use?

Currently the REQUIRE is set to 0.7. That unfortunately does not exist yet :). We can set it to 0.7-beta. Then femtocleaner will run as well, possibly catching anything you might have missed.

[jannefiluren]

Hi! Yes, that is done automatically using the Julia extension to VS code. I can easily do that on top of the new changes if you like. It takes a couple of minutes. So happily ignore #60 meanwhile.

I really like your work on this pkg. It is a very important one. Many thanks.

[meggart]

I think given the last changes I will not continue on this PR but will apply the list of changes described above in multiple smaller PRs. Will leave this one open anyway until there is a new PR.