New bba method

[JeanLucPons]

This PR adds a new BBA method (model dependent) that we use at ESRF.
Unfortunately it cannot be tested on design as the algorithm uses design so it needs a VA to be tested.

Example on ESRF VA:

from pyaml.accelerator import Accelerator
from pyaml.common.constants import Action
import numpy as np
import logging

def bba_callback(action: int, cdata):

    if action == Action.MEASURE:
        plane = cdata["plane"]
        step = cdata["step"]
        print(f"step {step} {plane} bpmpos = {cdata["bpm_pos"]*1e6:.3f} um")
    return True

logging.getLogger("pyaml.tuning_tools.bba2").setLevel(logging.DEBUG)

sr = Accelerator.load("tests/config/EBSOrbit.yaml")
SR = sr.live
bba = SR.get_bba("BBA2-BPM_C04-04")

# Add a misalignement
SR.get_bpm("BPM_C04-04").offset.set([20e-6,-15e-6])

bba.measure(sleep_between_step=6,callback=bba_callback)

print(f"HOffset: {bba.h_offset()}")
print(f"VOffset: {bba.v_offset()}")

outputs:

31 Jul 2026, 16:31:12 | DEBUG | Initial H corrector value: -9.66335494367311e-41 rad
31 Jul 2026, 16:31:12 | DEBUG | Initial V corrector value: 3.631830979621812e-43 rad
31 Jul 2026, 16:31:12 | DEBUG | Initial quad value: 1.779156783615382 m-1
31 Jul 2026, 16:31:12 | DEBUG | Moving in the negative direction: 0,0 -> -1e-05,-1e-05
step 0 H bpmpos = 68.273 um
step 0 V bpmpos = 17.329 um
31 Jul 2026, 16:31:26 | DEBUG | Moving in the positive direction: -1e-05,-1e-05 -> 1e-05,1e-05
step 1 H bpmpos = -28.900 um
step 1 V bpmpos = -47.134 um
31 Jul 2026, 16:31:38 | DEBUG | Moving to the best guess: 1e-05,1e-05 -> -1.194652000820969e-07,2.0562047474807868e-07
step 2 H bpmpos = 20.575 um
step 2 V bpmpos = -15.663 um
Moving to the optimum: -8.008225249053031e-08,1.3728543395643385e-07
HOffset: 2.038413409167885e-05
VOffset: -1.5442473918032936e-05

[gubaidulinvadim]

@kparasch maybe you'd like to have this kind of method directly in pySC? Model-dependent BBA is sometimes the only option.