Treat \empty as a table command

[lvjr]

Based on discussion #321, I think it would be good for users if tabularray treats \relax and \empty as table commands by default.

\NewTableCommand\empty{}
\NewTableCommand\relax{}

Update 1: Adding \NewTableCommand\empty{} is safe. But adding \NewTableCommand\relax{} causes an error even for a simple example:

> Step: build the whole table.

! Missing font identifier.
<to be read again> 
                   OT1/cmr/m/n/10 
l.11 \end
         {tblr}
I was looking for a control sequence whose
current meaning has been defined by \font.

Update 2: The error is due to the failure in restoring the original definition of \relax.

> Step: execute table commands.
> \relax=\protected macro:->\__cmd_start_expandable:nNNNNn
{}\__tblr_table_command_relax_saved:w  \__tblr_table_command_relax_saved:w 
\__tblr_table_command_relax_saved:w code ?{}.
> \empty=macro:->.

[lvjr]

Because \cs_if_exist:NTF doesn't treat \relax as an existing command (which makes it different from \ifdefined in eTeX), tabularray fails to save the correct definition of \relax.

\cs_if_exist:NTF #1
  {
    \cs_set_eq:cN { __tblr_table_command_ \cs_to_str:N #1 _saved:w } #1
  }
  {
    \exp_args:NcV \NewDocumentCommand
      { __tblr_table_command_ \cs_to_str:N #1 _saved:w } \l__tblr_a_tl { }
  }        

[lvjr]

I decided not to add \NewTableCommand\relax{} since it may be a little dangerous. But the problem of \cs_if_exist:NTF should be fixed so that users can add it if they need.