Remaining driver issues found while adding parameter value types

The issue lists affected driver behaviors, validators, helper annotations, and tests, but does not name their source file paths. Start by locating the named drivers and tests, then compare each validator or command with the manuals cited in the issue. Done means resolving the listed items with the needed compatibility or hardware checks; the issue spans many instruments and includes behavior changes that require domain knowledge.

While adding Parameter[<value type>, Self] annotations to the instrument drivers in #8658, a
number of driver bugs were found and fixed (see the PR and its newsfragments). The issues below
were found at the same time but are not fixed yet, mostly because they change behaviour or
validator ranges that we cannot verify without the instrument.

Missing parameter annotations are tracked separately and are not listed here.

Bugs

• Rigol DS1074Z: trigger_level in pattern mode does not send a channel. The
  MSO1000Z/DS1000Z Programming Guide (Dec. 2015) gives the pattern trigger level per channel:
  :TRIGger:PATTern:LEVel <chan>,<level> and :TRIGger:PATTern:LEVel? <chan>, e.g.
  :TRIGger:PATTern:LEVel CHANnel2,0.16. _get_trigger_level and _set_trigger_level build
  :TRIGger:<mode>:LEVel? / :TRIGger:<mode>:LEVel <value> from trigger_mode() for every
  mode, so no channel is sent in pattern mode. The other trigger modes (edge, pulse, video) are
  fine. A fix needs a way to choose the channel, e.g. a per-channel parameter.
• Lake Shore 325: wrong output_range label for loop 2. The Model 325 User's Manual
  (rev. 1.8) says the loop 2 RANGE is 0 = Off, 1 = On (single range voltage output, up to
  2 W), but the driver maps 1 to "Low (2.5W)", the loop 1 label. Renaming the value to
  "On" would break code that uses the old label, so this needs a deprecation path.

Non-integer values accepted for integer settings

These parameters read back an int but validate with Numbers, so a float passes validation.
Switching to Ints (as done for the Keysight PNA averages, points and active_trace in
#8658) would reject it, but is a behaviour change.

• CopperMountain M5xxx averages: Numbers(1, 999) with set_parser=int, so
  averages(2.5) silently sets 2.
• AimTTi PL-P curr_range: Numbers(1, 2), so curr_range(1.5) sends IRANGE<n> 1.5.
• Rigol DG1062 burst ncycles: Numbers(1, 500000), so a non-integer number of cycles is
  sent to the instrument.
• Keithley 7510 average_count: Numbers(1, 100), so a non-integer count is sent to the
  instrument.
• Lake Shore 372 channel dwell, pause, curve_number and excitation_range_number:
  validated with Numbers, so non-integer values are sent in the INSET / INTYPE commands.

Validator ranges that do not match the manuals

Not changed in #8658 since there was no instrument to test against.

• Agilent E8267C pulse_width: allows 10 to 20 ns. The Keysight E8257D/67D SCPI Command
  Reference (D model) gives 10 ns to (pulse period - 20 ns), and the *RST value of 1 us is
  outside the validator.
• R&S RTO1000 channel impedance: Ints(1, 100000), while the R&S RTO User Manual gives a
  real value from 0.1 to 100E+3 Ohm, so fractional values are rejected.
• Keithley 2450 nplc: allows 0.001 to 10, while the 2450 Reference Manual gives 0.01 to
  10.
• Rigol DG1062 (checked against the DG1000Z Programming Guide):
  • Burst ncycles stops at 500 000, the limit for internal trigger. With external or manual
    trigger the guide allows up to 1 000 000.
  • Burst time_delay has no upper limit, while the guide gives at most 100 s.
  • Burst idle accepts any number, while a user-defined idle level is an integer from 0 to
    16383.
  • Burst period starts at 3 us, while the guide's minimum is just over 2 us.
  • duty_cycle allows 1 to 99 %, while the guide gives 0.001 to 99.999 % for pulse waveforms
    and a frequency-dependent range for square waveforms.
• R&S ZNB bandwidth: the Enum stops at 1 MHz, but option R&S ZNB-K17 allows IF
  bandwidths up to 10 MHz.
• Keysight B1517A sweep:
  • power_compliance allows up to 80 W and current_compliance up to 40 A (the widest limits
    of all module types), while the B1500 Programming Guide gives Pcomp 0.001 to 2 W for the
    HRSMU (B1517A).
  • sweep_steps uses Ints(1, 1001), while WV allows 1 to 10001 steps.
• Lake Shore 325:
  • PID allows 0 to 1000 for P, I and D, while the manual gives 0.1 to 1000 for P and I and 0 to
    200 for D.
  • CRVHDR can only be set for user curves 21 to 35, but the driver allows setting the header of
    any curve.

Wrong type hints on helper functions

• Galil DMC-41x3: _set_relative_position, _set_speed, _set_acceleration and
  _set_deceleration are annotated val: str but receive int.
• QDev QDac: _getsync is annotated -> float but returns the int sync output number,
  which also prevents annotating sync as Parameter[int, Self].

Smaller issues

• Rigol DS1074Z trigger_mode has unit="V".
• Keysight 34980A: for a module without a driver, the fallback submodule name
  slot_<n>_<model>_no_driver includes the model string reported by the instrument. Model names
  containing - are not valid identifiers, which triggers a renaming UserWarning.
• Rigol DS4000 and Tektronix TPS2012 have no default_terminator, unlike most other VISA
  drivers. The tests pass terminator="\n"; whether the drivers should default to it needs
  checking on hardware.

Tests

• tests/drivers/test_agilent_e8257d.py: the hardware-only TestAgilentE8257D
  (DriverTestCase) uses a status parameter that the driver no longer has (it is now
  output_enabled), so it would fail if run against an instrument. It is always skipped in CI.
• tests/test_channels.py: the helpers _verify_multiparam_data and _verify_array_data are
  unused.