Improved stability and logging

Same as v0.4, with README.md updated

Consolidated the pipeline

- Three concurrent loops instead of two:
 - Tight loop: time integration
 - Medium loop: preprocessing
 - Loose loop: DMT / detection

-> Much less memory used

Fixed bug in acquisition, due to an unexpected packet being received.
Acquisition has been observed to run continuously without packet loss now.

Integrated the pipeline with the start_abort script at NenuFAR, for automatic launch.

Added the install target in CMake

First results on B1919

Starting to have an actually functional version:
Acquisition status: 75% functional
 - invalid memory access on long runs, not yet understood

Added preprocessing: 90% functional
 - Time integration only supported for FFTLEN=1 -> must add support for FFTLEN > 1
 - MAD Clean, Rm Baseline, Rm Mean functional

Dedispersion: 95% functional
 - Must check with N_t_in != N_t_out

Detection: 60% functional
 - Working version based on STD
 - Must switch / provide version with MAD (more robust statistic)
 - Must work on N_t_tmp values instead of N_t_out

Overall: 75% functional
 - Bounded by acquisition
 - Need for multiple time integration factors