data.table optimization of SPC j-index and adding k-keywords (#201)

* Use data.table for all j-index subsetting #157

* Significant optimization for j-index subsetting of SPC

* Add more competitive version of base function for j-index

* spc_j_base_2: 100k benchmark ~0.1s faster than original, nearly 3x slower than new DT

* spc_j_base_3: more competitive for small SPCs

* add demo of SPC non-standard eval "k-index" .LAST keyword #199 #201

* add experimental support for .FIRST, .HZID and multiple special keywords

* better logic for i+j+k

* Convert demo to tests

* Docs