Fix stray control bytes tearing down the ASTM connection

[xispa]

Description of the issue/feature this PR addresses

Follow-up to #77 (chunked-frame reassembly). Raising inside an asyncio protocol's data_received() is fatal -- it drops the connection. Several ASTM handlers raised on input that instruments legitimately send, so a single stray byte forced a disconnect/reconnect even when framing was fine:

• on_eot() raised InvalidState when an <EOT> arrived outside a transfer, but instruments emit stray <EOT> (and <EOT>/<ENQ> bursts) between sessions and while establishing.
• on_ack()/on_nak() raised NotAccepted on any <ACK>/<NAK>; as the receiver we should ignore these.

Observed live against an Abbott ARCHITECT ci4100: after #77 made frames reassemble correctly, <EOT>/<ENQ> bursts between sessions still crashed the connection repeatedly (disconnect + reconnect churn).

Current behavior before PR

A stray <EOT> outside a transfer, or any <ACK>/<NAK>, raises inside data_received() and tears down the connection, forcing a reconnect. One unexpected byte can interrupt an otherwise healthy link.

Desired behavior after PR is merged

Stray/unexpected control bytes are non-fatal: on_eot() outside a transfer and unexpected on_ack()/on_nak() log-and-ignore, and the unit-dispatch loop wraps handle_data() in try/except so no single byte can drop the link. The stray-<EOT> branch does not call reset_session_state() (that would clear the buffer mid-dispatch and drop bytes following the EOT in the same read). Adds tests for stray <EOT>, <EOT>/<ENQ> bursts, an <EOT>+<ENQ> in one read, and unexpected <ACK>/<NAK>.

--
I confirm I have tested this PR thoroughly and coded it according to PEP8
and Plone's Python styleguide standards.