try_inverse and ownership

Hi, I noticed in the [docs for `try_inverse_mut`](https://docs.rs/nalgebra/latest/nalgebra/base/struct.Matrix.html#method.try_inverse_mut):

> If you need to preserve the original matrix regardless of success or failure, consider using [Self::try_inverse](https://docs.rs/nalgebra/latest/nalgebra/base/struct.Matrix.html#method.try_inverse) instead.

But `try_inverse` takes `self`, not `&self`, and consumes the matrix. So, two points:

1. Should the docs be reworded? To me, the implication that `try_inverse` preserves the original matrix seems to be false. Or is something else meant by "preserved"?
2. Is there a possible performance optimization here, to make`try_inverse` not take ownership of the matrix, and thus allow the end-user to avoid `.clone()`? We still have to allocate memory for the LU decomposition, but we can at least save CPU cycles copying all the values, right?