Define CS-specific behavior of algorithms only with strategy (part 2).

[awulkiew]

Followup of #533

The changes:

• rtree may take a strategy as bgi::parameters<RtreeParams, Strategy>, not all possible operations are supported, only those required by other algorithms where the rtree is used, for other operations the default strategy is used as it was before. This needs further changes but I think it needs more general changes in strategies, i.e. introduction of "proper" umbrella strategies.
• rescale_policy_type takes optional CSTag (previously CS-related behavior was defined by Geometry type).
• the information about the CS taken from the strategy is propagated internally in various algorithms.
• Point types are now optional in point_in_poly strategies (left as template arguments defaulted to void for backward compatibility).
• getters for CS-specific strategies added to some strategies.

[vissarion]

Thanks for this PR. I am ok with merging.

[vissarion]

why not typedefed as above or not typedef the above to look alike. Similar cases below. Am I missing something?

[awulkiew]

No, there is no specific cause. I think I prefer removing typedef above because it's used only in one place.

[barendgehrels]

Thanks! It is huge, and I could not check all details, but what I've seen looks to me. So I'm OK with merging this.

[awulkiew]

Thanks, I think I'll also add some tests for geometries with undefined CS before merging.

[awulkiew]

I addressed the issue raised by @vissarion, resolved the conflict with develop, added tests and fixed one more issue, i.e. passed the strategy into get_rescale_policy() which calls envelope() and expand(). I plan to merge it this weekend.