Fix entry addition logic for keyless match tables

[Shivampal157]

Summary

Fixes entry addition logic for keyless match tables.

Previously, adding an entry to a table with no match key (get_size_key() == 0)
would always fail with NO_TABLE_KEY, even when the provided match_key was empty.

What changed

• NO_TABLE_KEY is now returned only when:
  • the table has no match key, and
  • a non-empty match_key is provided
• Keyless tables with a single default action can successfully add entries again.

Files changed

• src/bm_sim/match_units.cpp

Motivation

This restores correct behavior for keyless tables such as get_lb_key,
which rely on a default action without match keys.

[jafingerhut]

I do not understand why you believe it should be possible to add an entry to a P4 table with no key fields.

If you never add an entry to such a table, it should always get a miss on any table_name.apply() call, and execute the table's default_action. This is what happens today, yes?

Said another way: suppose it was possible to add an entry to such a keyless table. Should future calls to table_name.apply() then get a hit, instead of a miss? Why?