Allow 'type.convert' argument in tstrsplit function to support a function/list of functions/named list.

[Kamgang-B]

Closes #5094.
In addition to the current behavior of the tstrsplit function, I suggest to extend its features by allowing its 'type.convert' argument
to support a function/list of functions/named list (where its names are functions).
In the case of a named list, the last element may not be named if it is a function.

New features suggested (The pieces of code in the list below are assumed to be the specification of the argument type.convert of the function tstrsplit.):
if:
1. type.convert=fun, then the function fun applies to each element of the transpose list.
2. type.convert=list(fun), --> same behavior.
3. type.convert=list(fun_1, fun_2, ..., fun_n) and the transpose list has length n, then fun_1 applies to the first element, fun_2 to the second element, etc.
4. type.convert=list(fun_1=c(1L, 3L)), then fun_1 applies to the first and third elements (and the remaining elements are kept as is; that is, as character vectors).
5. type.convert=list(fun_1=c(1L, 3L), fun_2=4:5), then fun_1 applies to the first and third elements, and fun_2 applies to the fourth and fifth elements (and the remaining elements are kept as is).
6. type.convert=list(fun_1=c(1L, 3L), fun_2=4:5, fun_3), --> same but the fun_3 applies to the remaining elements; so they are not kept as is.

Any new behavior different from those above is likely unexpected.

Any critic is welcome.

I can also extend or shrink the features above in case there is a need.

To show how these new features work in practice, here are some illustrative examples:

library(data.table)
options(datatable.print.class=TRUE)

DT = data.table(
  w = c("Yes/F", "No/M"),
  x = c("Yes 2000-03-01 A/T", "No 2000-04-01 E/R"),
  y = c("1/1/2", "2/5/2.5"),
  z = c("Yes/1/2", "No/5/3.5"),
  v = c("Yes 10 30.5 2000-03-01 A/T", "No 20 10.2 2000-04-01 E/R"))

#         w                    x           y          z                            v
#    <char>               <char>      <char>     <char>                       <char>
# 1:  Yes/F   Yes 2000-03-01 A/T       1/1/2    Yes/1/2   Yes 10 30.5 2000-03-01 A/T
# 2:   No/M    No 2000-04-01 E/R     2/5/2.5   No/5/3.5    No 20 10.2 2000-04-01 E/R

# convert each element in the transpose list to type numeric
DT[, tstrsplit(y, "/", type.convert=as.numeric)] 
#       V1    V2    V3
#    <num> <num> <num>
# 1:     1     1   2.0
# 2:     2     5   2.5

# convert each element in the transpose list to type factor
DT[, tstrsplit(w, "/", type.convert=as.factor)]
#        V1     V2
#    <fctr> <fctr>
# 1:    Yes      F
# 2:     No      M
DT[, tstrsplit(w, "/", type.convert=list(as.factor))]  # same

# can 'also' convert all character vectors to factors
DT[, tstrsplit(z, "/", type.convert=function(x) type.convert(x, as.is=FALSE))]
#        V1    V2    V3
#    <fctr> <int> <num>
# 1:    Yes     1   2.0
# 2:     No     5   3.5

# convert part (some elements) of the transpose list and leave another part (if there is any) unchanged.
DT[, tstrsplit(z, "/", type.convert=list(as.numeric=2:3))]
#        V1    V2    V3
#    <char> <num> <num>
# 1:    Yes     1   2.0
# 2:     No     5   3.5

# convert some elements to the corrisponding specified types and the remaining elements to a given type
DT[, tstrsplit(z, "/", type.convert=list(as.factor, as.integer, as.numeric))]          
#        V1    V2    V3
#    <fctr> <int> <num>
# 1:    Yes     1   2.0
# 2:     No     5   3.5

DT[, tstrsplit(z, "/", type.convert=list(as.factor=1L, as.integer=2L, as.numeric=3L))]
#        V1    V2    V3
#    <fctr> <int> <num>
# 1:    Yes     1   2.0
# 2:     No     5   3.5

# convert some elements to the corrisponding specified types and the remaining elements to a given type
DT[, tstrsplit(z, "/", type.convert=list(as.factor=1L, as.numeric))]
#        V1    V2    V3
#    <fctr> <num> <num>
# 1:    Yes     1   2.0
# 2:     No     5   3.5

# convert given elements to specific types and convert the remaining elements using 'type.convert' function.
DT[, tstrsplit(v, " ", type.convert=list(as.factor=1L, as.IDate=4L, function(x) type.convert(x, as.is=TRUE)))]
#        V1    V2    V3         V4     V5
#    <fctr> <int> <num>     <IDat> <char>
# 1:    Yes    10  30.5 2000-03-01    A/T
# 2:     No    20  10.2 2000-04-01    E/R

DT[, tstrsplit(w, "/", type.convert=list(as.factor=1:2, as.numeric))]
#        V1     V2
#    <fctr> <fctr>
# 1:    Yes      F
# 2:     No      M
# Warning message:
# In the argument 'type.convert', 'as.numeric' was ignored because all elements in the transpose list or elements corrisponding to indices specified in the 'keep' argument have already been converted. 

# errors: 
DT[, tstrsplit(z, "/", type.convert=list(as.factor, as.numeric))]      #  2-length type.convert vs 3-length output
DT[, tstrsplit(z, "/", type.convert=list(as.integer=2L), keep=5L)]     #  type.convert not contained in the keep argument
DT[, tstrsplit(w, "/", keep=integer())]                                # keep empty. CURRENT BEHAVIOR OF tstrsplit returns empty list with warnings.
DT[, tstrsplit(w, "/", type.convert=list())]                           # empty list non supported     

[mattdowle]

LGTM. Great PR. Thanks.
Have invited you to be project member so among other things you can create branches in the main project now rather than in a fork. The invite should be a button you need to click to accept in your GitHub profile or projects page.
If you'd like your surname to be in DESCRIPTION (displayed on CRAN page here: https://cran.r-project.org/web/packages/data.table/index.html) instead of just "B" please let me know what it is.

[Kamgang-B]

Hi @mattdowle
Yes, I prefer my full name (if possible).
name: Boniface Christian
Family name: Kamgang
person("Boniface Christian", "Kamgang")