~10x speedup for webform_submissions

@podarok - Thank you very much for your efforts at optimizing webform. In reading your patch, I have a few comments:

1. This patch would only be considered for the 7.x-4.x and 8.x branches. A back-port to 7.x-3.x is possible, since we have a new maintainer for that branch.
2. The patch doesn't apply as I believe it isn't made at the right directory level.
3. Your patch spans several completely unrelated optimizations. I'd like to see each presented individually. I'd also like to see some quantitative data (workload, timing results with and without the optimization). Obviously we only want to apply optimizations that are effective.
4. The optimization to webform_component_feature, if effective, looks good but could be more easily made with the static design pattern. There would be no need for another module to reset the cache at runtime, and the features could simply be stored in one static array. This mean that webform_components() would be called exactly once, rather than once for each combination of $type and $feature, resulting in a simpler and more effective optimization.
5. The optimization to webform_results_export / $submission_information is optimizing an array of only 3 value and if it makes even a nanosecond of difference then my head will explode at the awfulness of PHP. The foreach should call array_values() exactly once.
6. There are some line spacing and whitespace changes that should be removed. They just complicate the patch.
7. I haven't studied the pager code carefully, so my comments are just things that I'm wondering. Are the table references 'ws.' actually needed to resolve ambiguity? If not, they should be removed as a stylistic difference.
8. I'm surprise to see the call to fetchAll(). I would have thought that this would require significantly more memory. Have you profiled this change individually?

Please don't be deterred. I am very enthusiastic about committing effective optimizations. Also, I've adjusted the issue metadata to correspond to Issue Queue Handbook standards.

@podarok -- As luck would have it, user torotil has just volunteered to be webform's new 7.x-3.x branch maintainer. The good news is that since both you and he are interested in maintaining this branch, you can review each other's work. He has another issue in the queue that I'm sure he'd like eyes on (#1332820: Parts of nested tokens (in fieldsets) printed when left empty). And I discussed this issue with him and he's ready to help review patches after the review in #5 is attended to.

I'm happy to see the enthusiasm for 7.x-3.x and for optimization.

No activity in a month, so I went through this patch again in the context of the active branches.

1) I see static caching added to webform_component_feature(). It doesn't do much, except call webform_components(), which is itself cached. I'm unconvinced of a big win here, but i didn't do any testing to confirm this. I'd be very happy to be proved wrong.

I did make a tiny optimization to webform_components, eliminating a array_diff_key. With only a dozen component types, I doubt it helps much, but I think it makes the code clearer.

2) There is a probably big win optimization to webform_get_submissions when a header is used. Due to changes in the 4.x branch, webform_get_submissions is no longer called for the submissions or results table, as these are done with views. The 8.x branch should remove the header option probably.

3) I don't understand how the change to webform_get_submission_count would help. I didn't check the 3.x branch, but the 4.x branch has indices on webform_submissions to make queries with nid and optionally uid very fast. I did see a line of left-over code, which I deleted.

Committed to 7.x-4.x and 8.x.

Alas the big win is an optimization in code no longer used in the current branch.

My thanks to @podarok for the webform_get_submissions optimization. It will end up in webform as part of the optimization and fixes for batch download. See the related issue.