D270 - Real hardware is a pointer from the compatibility table, not a column in it

docs/COMPATIBILITY.md compares loaders - host, shadPS4, PS5PCEM, fpPS4, kyty, orbistoun - and a reader reasonably asks where the real PS5 is. It is not a column, deliberately.

Two reasons. First, a retail console is not a loader being compared; it is the source of truth the loaders are scaffolding toward (the document's own framing), so ranking it in the same table as the approximations inverts what the table is for. Second, the compat table is generated by obscene-tool compat from emulator report files and gated against drift by verify.sh (D069). A hardware tally added as a column would either have to be regenerated from a committed hardware report on every run - coupling a gated, machine-maintained table to a measurement that changes shape as the suite grows - or be hand-maintained beside machine-maintained cells, which is the exact staleness this project quarantines patched-loader results for (D176).

So the hardware results live in docs/HARDWARE.md, cited record by record to data/hardware/ (ps5-full.txt before imports bound, ps5-imports.txt after), and COMPATIBILITY.md carries a prominent pointer section to them rather than a copy. The committed screenshot screenshots/ps5-hardware.png is shown in both. ./bin/obscene report / deploy capture a fresh run (D269). This is what the question "why is there no hardware column?" resolves to, recorded so the answer is not re-derived as a column that breaks the gate.

Status: done.