Add `[.icd9cm_pc` and `[.icd10cm_pc` methods

[dapritchard]

Since extraction methods aren't defined for icd9cm_pc and icd10cm_pc classes, extraction resorts to the default method which drops all of the type encoding information. See the following for an example.

library(icd)
z = as.short_diag(as.icd9cm_pc(c("0053", "0053", "0211")))

z
#> [1] "0053" "0053" "0211"
#> attr(,"class")
#> [1] "icd9cm_pc" "character"
#> attr(,"icd_short_diag")
#> [1] TRUE

z[1L]
#> [1] "0053"