transfer of the record length for io from internal files

[THOMAS Paul Richard 169137]

By the way, I did not intend that this patch be committed.

I just wanted to establish:

1) Is ioparm.recl_in the correct member to use? It is appropriately named
and is not use, otherwise, in internal io.

2) What should be the structure of this addition to trans-io.c?  Looking at
it in the cold light of day, I would propose to revert to set_string, as it
was, and to add an extra function, set_internal_file that contains the few
extra lines of code and calls set_string.

3) Has anybody checked that all the string parameters that must be scalar
are asserted to be so in the parser?  Maybe the cleanest way of dealing with
this would be to write a third function, set_scalar_string, that contains an
assertion that this is the case?  It's a bit belt-and-braces but would
likely look better in the code and provides a backstop, just in case.

4) A means for Jerry to develop the library part.  Should I, to that end,
clean up the patch and submit it to be committed?  As the joke says, it does
no harm.....

Paul T