clean code and docstrings + improve performance

[heliosdrm]

AbstractRecurrenceMatrix and textrecurrenceplot are given in the documentation, so they should be exported.

Since the docstring of rqa goes first in the documentation pages, I have moved the full explanations of the behavior and conditional defaults of theiler there, and left only a reference to rqa in the rest of functions. (This also makes the text less repetitive.)

Explanations added about the ambiguities in the literature about the definition of recurrence times (in recurrencestructures) and the denominator of the recurrence rate (in recurrencerate), with a rationale of the choices that we have made in this package.

[heliosdrm]

The docstrings of the individual RQA functions are now expanded with more details about how they are computed (I hope I have not made mistakes with the latex formulas)

[Datseris]

Awesome! I was going to open an issue for more docs in the RQA functions after we had 1.0 out, to not delay the release, but you did it already.

I will review this properly by building the latex documentation and checking all latex expressions when I am back home late night!

[heliosdrm]

Made a change in trend which is not only documentation, but to gain performance.

That change makes the result slightly different for CrossRecurrenceMatrix made with time series of different lengths. But both results are consistent with the published formula, and I have not found results that I can compare in other software packages or fabricated examples. So perhaps we can set our own standard for such a rare case...

Or if we find such standard in the future and realize that the current computation is wrong for such cases, we fix the bug then.