Additional nanoarrow refactoring

Building on #663 (and its SC 41534), additional clean up of the Arrow import/export code and decrufting is done here.

All interface changes are at the mostly internal function level and concern the switch from operating on an explicit two-element list of Arrow Array and Schema external pointers to an (nanoarrow-special) ArrowArrray external pointer which tugs the Schema external pointer into the tag field of itself (ie the Array pointer) making the function single argument (nicer with C API of R we have to adhere to) and a SEXP itself (as an external pointer is a possible SEXP instance within the scope of the SEXP union type). Cleverly the SEXP is actually classed as an S3 object by nanoarrow so after eg creating quickstart_sparse we can do this:

> library(tiledb) 
TileDB R 0.25.0.3 with TileDB Embedded 2.22.0 on Ubuntu 23.10.  
See https://tiledb.com for more information about TileDB. 
> library(nanoarrow)   
> uri <- "quickstart_sparse" 
> arr <- tiledb_array(uri, "READ") 
> qry <- tiledb_query(arr, "READ") 
> values <- integer(8)       # R shortcut to reserver a buffer
> tiledb_query_set_buffer(qry, "a", values) 
> tiledb_query_submit(qry) 
> tiledb_query_finalize(qry) 
>
> # we now get a nanoarrow S3 object which has a print method
> na <- tiledb_query_export_buffer(qry, "a") 
> na 
<nanoarrow_array int32[4]>
 $ length    : int 4  
 $ null_count: int 0 
 $ offset    : int 0     
 $ buffers   :List of 2 
  ..$ :<nanoarrow_buffer validity<bool>[0][0 b]> `` 
  ..$ :<nanoarrow_buffer data<int32>[4][16 b]> `1 2 3 2`
 $ dictionary: NULL 
 $ children  : list()
>
> # and arrow converters
> arrow::arrow_array(na)
Array
<int32>
[
  1,
  2,
  3,
  2
]
>

TileDB-Inc / TileDB-R

Additional nanoarrow refactoring #682