det: LINK RESP2, RETURN TRANSID LENGTH past the DTO, and onward pass-through of opaque COMMAREA bytes (follow-up to #4501) #4679

Left by #4501 (PR #4678), which passes an XCTL's over-DTO COMMAREA bytes as opaque bytes:

• A failed LINK still stores RESP2 0 in the det port (outcome(..., "0")); XCTL now derives it (DetCics.xctlResp2: LENGERR 11, PGMIDERR 1). LINK needs the same, or CicsTask returning RESP2 (det runtime: browse/DELETE RESP2 as IBM states it; signon model port LOW-VALUES (#4657, #4659) #4661).
• RETURN TRANSID COMMAREA LENGTH past its DTO still passes the DTO (CardDemo: 2000 over 172 bytes); passing bytes needs equivalence_cics.from_java to accept a byte[] COMMAREA (equivalence: CICS proofs compare DISPLAY output and keep low-values apart from spaces (#4635) #4660 area).
• Bytes past the receiver's own DFHCOMMAREA are dropped, so a further RETURN/XCTL does not pass them on; none of the corpus cases needs it yet.
  Part of COBOL translator: CICS command breadth (START/RETRIEVE, channels/containers, GETMAIN, LOAD, CONVERSE, INQUIRE, ASSIGN, counters, HANDLE ABEND PROGRAM) #4270.