Remove webpack-directory-name-as-main dependency and corresponding non-standard default module behavior

[epixa]

There are a bunch of modules that are relying on a custom loader in
webpack to change the expected loading scheme for modules so that a file
with the same name within a module would be treated as a default for
that module. In order to ultimately remove that custom loader, we add
default index.js files that proxy to those existing module defaults.

This is the safest option, as it preserves both the default loading
behavior for the module as well as any direct module traversals to the
real file.

The custom loader is also removed entirely from the project.

CommonJS syntax is used for consistency's sake: some of the modules that
are updated are webpackShims, and they cannot use es6 syntax.

[epixa]

@tylersmalley I seem to remember you saying that you wanted to see this gone.

[tylersmalley]

jenkins, test it

[tylersmalley]

The test failure:

ERROR in ./src/ui/public/test_harness/test_harness.js
    Module not found: Error: Cannot resolve 'file' or 'directory' /home/jenkins/workspace/kibana_core_pr/src/ui/public/stack_trace_mapper in /home/jenkins/workspace/kibana_core_pr/src/ui/public/test_harness
     @ ./src/ui/public/test_harness/test_harness.js 24:28-60

[tylersmalley]

Shouldn't we be able to actually remove most of the files, like ace? They are required within the application and not actually a shim.

[spalger]

I fully support removing this behavior, and regret introducing it to being with, but I really don't think we should spread the use of common js. The webpack shims are defined for 3rd part code, which is why they don't use es6, but that doesn't mean we are going to use es5 everywhere..

[epixa]

I agree.

[epixa]

I'm going to close this PR since this sort of change is done with automation, so I'd need to apply it again to the whole project anyway at this point.

I could never get this build to pass and I'm not quite sure why. The root cause was buried in a lot of indirection pretty deep inside certain abstractions, so I'll need to figure that one out before issuing a new PR.