There are two important structures. RTreeNode and RTree. There are 4
basic functions to interact with this structures: create and rtree,
insert something to it, query and rtree and free the memory of an rtree.

We include a basic structure for the Rtree which is working for STBoxes.
It allows to insert one by one. Queries are done with an STBox. A
function to find each value as a double is required to be passed down to
the RTree itself.

Since there is no tests for C code and only for sql code, this example
is in some way also a test.

It shows how you would normally use the Rtree, inserting and querying.

The largest axis was always choosing to split via the first element. Now
it changes. Some thought should be done for the temporal axis since it
is normally in a different scale. Maybe a mahalanobis distance would be
better.

meos.h should only have information that the user would use. RTreeNode
is no an structure that a user would use. Thus, is better to have it in
tpoint_rtree.h and have a forward definition in meos.h for RTree node.

Keeping it consistent with the rest of the codebase

RtreeNode will allocate only once and there will be no need to allocate
when inserting since it copying to an already allocated block of memory.

I have assumed meostype as an int since I meostype is defined in
meos_catalog and I can't import it from meos.h, thus int was best option
I had.

It was ambiguous that both #defines were related so they were refactored
as RTREE_INNER_NODE and RTREE_INNER_NODE_NO

Adds missing documentation to newly add functions

To answer the query of rtree_search it is necessary to realloc memory,
thus the code can become a little obscure. Added one extra function and
some comments to make life easier for maintenance.

Included some time taking sinceit is important to know the difference
between bruteforce and RTree search.

Use a simple STBox instead of a palloc.