[BREAKING] deprecate names=true in eachcol

[bkamins]

Fixes #2090.

In the next move we will drop V parameter of DataFrameColumns.

Making this PR revealed that names=true is in practice almost never used (at least internally), so this change - while big - should not be that problematic.

[bkamins]

I decided to redesign eachcol in one shot. What will be breaking:

• eachcol(df, true) will print differently (it will be a plain array)
• eachcol(df, true) will not have getproperty and propertynames defined

Given that probably eachcol(df, true) is not used that much and we introduced getproperty only recently I think it should be OK to go this way.

[quinnj]

There was a discussion the other day on slack that mentioned a convention of adding [BREAKING] to PRs that introduce breaking changes. This can help later when you review tagbot release notes w/ PRs that start w/ [BREAKING].

[bkamins]

We now use a label breaking for this. But I can also add [BREAKING] in the name

[bkamins]

@nalimilan the PR is good to have a look at. the ❌ is due to coveralls only.

[bkamins]

❌ is due to coverage

[bkamins]

The failing build on nightly is due to JuliaLang/julia#34889 so I will merge this.

[pablosanjose]

I just made a PR with a fix for JuliaLang/#34889, sorry about that.

[bkamins]

@pablosanjose - thank you for such a quick fix.