Accept the optional timer byte in 0x1946

[smoke0leg]

What

0x1946 (Control Floor Heating Status) is encoded as a fixed 7-byte payload.
Some DLP thermostats reject that length and require an 8th byte — the timer —
which is the same field 0x1945 already handles as optional.

This makes 0x1946 symmetric with 0x1945: parse reads timer when the
payload is at least 8 bytes, encode appends it only when data.timer is set.
Payloads without a timer are byte-for-byte unchanged, so the change is
backwards compatible.

How it was found

Measured on a live Buspro installation (9 devices on subnet 1, three 0x23AB
DLP floor-heating thermostats). Controlling the floor heating failed
consistently until the payload length was changed:

Payload	0x1947 response
7 bytes	F5 — failure
8 bytes, timer appended	F8 — success

Same device, same field values; only the length differs. HDL's own
documentation is not public, so this is behaviour observed on the wire, not a
spec citation.

Tests

Added a second fixture for 0x1946 — one example with the timer and one
without — mirroring the pair that already exists for 0x1945.

The test was checked against the bug it claims to catch: with the new fixture
in place but lib/commands.js reverted, the suite fails with

AssertionError: expected Buffer [ 00, 01, 01, 16, 16, 14, 12 ]
      to equal Buffer [ 00, 01, 01, 16, 16, 14, 12, 01 ]
      (at length, A has 7 and B has 8)

With the fix applied: 207 passing.

npx eslint lib test reports the same 13 problems before and after this change
(all pre-existing and in unrelated lines), so I left them alone to keep the
diff focused.

Not included

0x1947, the control response, may carry the same optional byte. I have no
capture proving it, so I did not touch it.

🤖 Generated with Claude Code