Adds the option to calculate sequence waveforms by multiple processes

[schote]

• Splits waveform calculation and sequence provider into two modules
• Adds two-staged sequence calculation: After sequence was initialized, digital control signals are added sequentially (fast). In a second step the heavy computation of all the waveforms is done. This can be done sequentially or by using multiple processes.
• Uses np.memmap to store the calculated sequence waveforms

To do:

• Compare np.memmap and shared memory

[github-actions]

Coverage Report

File	Stmts	Miss	Cover	Missing
src/console
__init__.py	10	1	90%	11
src/console/interfaces
acquisition_data.py	123	25	80%	96–99, 110–111, 129–130, 145–148, 158–159, 188–189, 193–199, 215–216, 227, 231, 249
acquisition_parameter.py	107	6	94%	126, 147–148, 163, 178, 213
device_configuration.py	55	2	96%	61, 67
dimensions.py	96	25	74%	45–46, 67–68, 71, 74, 78–80, 89, 95, 101, 107, 113–115, 125–127, 137–139, 149–151
rx_data.py	86	18	79%	77–81, 89, 101, 105, 107, 122, 130–135, 143, 151, 154, 157
src/console/pulseq_interpreter
_block_calculator_backup.py	81	81	0%	3–266
block_calculator.py	113	11	90%	49, 51, 59, 92, 111, 123, 129–130, 239–241
sequence_provider.py	220	83	62%	84, 141, 164, 174–178, 180–181, 187–190, 206–209, 212, 236–238, 294–336, 352–354, 358–360, 366–367, 409, 412–421, 426–430, 435–438, 446–447, 449–450, 455, 457–458, 465, 469
src/console/spcm_control
abstract_device.py	79	79	0%	2–149
acquisition_control.py	148	148	0%	3–362
rx_device.py	209	209	0%	2–493
tx_device.py	169	169	0%	3–431
src/console/spcm_control/spcm
pyspcm.py	161	161	0%	3–295
tools.py	54	54	0%	3–128
src/console/utilities
json_encoder.py	8	3	62%	22–24
load_configuration.py	20	6	70%	18–20, 28–30
plot.py	53	16	70%	31–32, 80–98
src/console/utilities/data
mrd_helper.py	43	8	81%	37, 43, 49, 52, 60, 63, 68, 71
write_acquisition_to_mrd.py	49	5	90%	43–44, 68, 86, 91
src/console/utilities/sequences/calibration
fid_tx_adjust.py	22	1	95%	66
se_tx_adjust.py	29	2	93%	62–64
src/console/utilities/sequences/spectrometry
fid.py	13	13	0%	2–74
se_spectrum.py	26	2	92%	53, 109
src/console/utilities/sequences/tse
tse_3d.py	242	75	69%	110–111, 113–114, 116–117, 124–125, 127–129, 131–134, 145–148, 153–156, 163–174, 191, 213–220, 226, 257, 320–337, 344–345, 412, 415–427, 535–546
TOTAL	5474	1203	78%

Tests	Skipped	Failures	Errors	Time
180	0 💤	0 ❌	1 🔥	46.114s ⏱️

[schote]

@Toreil I noticed that the adc dead times in the adc_library and the adc event returned by get_block do not necessarily align. I think I would rather trust the dead time defined in the adc_block. See imr-framework/pypulseq#365

[schote]

A quick measure of sequence calculation time over number of processes shows the expected behavior:

Code snippet:

exec_times = []
processes = [1, 2, 4, 8, 16]
for num_proc in processes:
    with AcquisitionControlManager() as m:
        start_time = time.time()
        m.acquisition.set_sequence(sequence=seq, parameter=console.parameter, num_processes=num_proc)
        exec_times.append(time.time() - start_time)

[schote]

This PR already contains #159 and #160. Therefore, these PRs should be merged first.