see commit 622b919b

TODO: if the UE did random access (followed by a MAC uplink with
CRNTI) because none of its scheduling request was granted, then
according to 36.321 5.4.4 the UE's MAC will notify RRC to release
PUCCH/SRS. According to 36.331 5.3.13 the UE will then apply
default configuration for CQI reporting and scheduling requests,
which basically means that the CQI requests won't work anymore and
that the UE won't do any scheduling request anymore as long as the
eNB doesn't reconfigure the UE.
We have to take care of this. As the code is, nothing is done and
the UE state in the eNB is wrong.

removed "short_offset" code from normal_prefix_mod. removed static declaration in PHY_ofdm_mod which was a cause for problems when parallelizing FFTs for FEP TX.

- change target values
- change upper/lower limit to trigger a tpc
- don't use ul_cqi from SR

The value of ul_cqi is not convincing, for both PUSCH and PUCCH,
more work/analysis is required.

put pucch metric into uci->stat before calling harq_indication. Fixes an issue with PUCCH1 SNR measurements coming from HARQ.

added deactivation of ULSCH harq process after 3rd retransmission. Avoids problem of accumulating failed RA processes.

Adjustment for received power calculation in eNB pusch procedures. Added factor to account for mcs (mimic the power control adjustment in UE).
activated periodic trace for PHR/PUSCH SNR/PUCCH SNR in MAC.

- better ulsch harq management in phy (maybe not over)
- don't do schedule_CSI if UE not in RRC_CONNECTED (not sure if correct)
- don't call find_ulsch in fill_dci0
- some spaces cleaned
- log improved

this is not handled correctly as of now, let's crash

-1 does not work

- re-use ta_timer
- when ta_timer is running:
  - no TA sent to UE
  - no TA from UE (for UE to have time to apply previous TA command)
- decrease ta_timer every TTI
- be careful to use correct value of ta_update (31 means no TA)

Maybe not correct, to be checked.