Fixed a mistake in the C++ code.

- The main function
`gfilmm`

now has an option`nthreads`

, to run some pieces of code of the algorithm in parallel. - Use of a sparse matrix to reduce memory consumption and to improve the speed.
- New function
`gfiQuantile`

.

- Removed the quadruple precision feature because of issues with the CRAN checks.
- The main function
`gfilmm`

now has an option`long`

, to run the algorithm in long double precision instead of double precision.

- Fixed two mistakes in the C++ code.
- Fixed a mistake in the R code; the results were wrong when the model had random effects with interaction.
- The main function
`gfilmm`

now has an option`precision`

, to run the algorithm in long double precision or quadruple precision instead of double precision. - It also has a new option seed, to set the seed of the C++ random engine.
- Updated the vignette.

First release.