This reverts commit d234366d.

This reverts commit 4f9f1720.

with previous updates in cyber docker file, rebuild cyber docker image.

This reverts commit e2f94a1e.

* cyber: lay the ground work of docker scripts for adding arm support

add arm support for Cyber RT docker containers.

Due to the problem with docker exec on Drive PX platform,
the start and into process is a little bit different.
Have added enough print out information from scripts to help.

Here is the typical process:

First time, start and getting into the container:
  ./docker/scripts/cyber_start.sh
  ./docker/scripts/cyber_into.sh
then run:
  /apollo/scritps/docker_adduser.sh
  su nvidia

After the first time, you can just use the into script to
get into the container:
  ./docker/scripts/cyber_into.sh

To exit, please use "Ctrl+P" then "Ctrl+Q". Otherwise, you
will stop the container and have to restart as shown for
the first time instructions.

* cyber: prevent LD_PRELOAD for cyber only development

Apollo add the default LD_PRELOAD for caffe and opencv which is not
available in Cyber RT only docker environment, so to prevent excessive
warnings, we reset the LD_PRELOAD when necessary.

* cyber: modify core cyber files to make build work on arm

todo: make build work on both x86 and arm

* cyber: croutine support for arm

* cyber: make swap build works on both x86 and arm

* cyber: fix croutine marcos

* drivers: fixed build error on float/double

* cyber: llround has different namespace on arm

* cyber: fix signed and unsigned comparison warning

* drivers: make bcan works on both x86 and arm

* drivers: adapt camera drivers on arm platform

* add CPU version of yuv to rgb conversion
* remove Apollo camera triggering tool dependencies

* cyber: make camera driver build works on both x86 and arm

* drivers: fix lint problem.

* Tools: Add example job config.

* Tools: Add job submission doc.

* Tools: Add BOS region in example config.

* planning: move common code out, to add more functions use shared code

* Dreamview: fixed control curvature calculation

* Prediction: implement Clear function in pedestrian interaction evaluator

* Dreamview: define js-yarml version for security purpose

* Prediction: load models

* Prediction: add checkout for ego_feature on semantic_map

* Prediction: separate thread_pool for caution level obstacles

* Prediction: framework of pedestrian interaction evaluator

* Prediction: partially implement pedestrian interaction evaluate func

* Planning: fix trajectory partition

* Planning: use traj history in CheckTrajTraversed

* Planning: move a few parameters to GFLAGS.

* Planning: add lateral and longitudinal offset check to gear change point

* Planning: increase resolution of iterative anchoring smoother

* add open space obstacle buffer in pull over scenario

* Planning: switch to iterative anchoring smoother and do replanning time only when fallback for testing

* planning: fail pull-over sonner based on path_data, and add a temp stop fence so that adc can be paused at a better position

* Planning: slightly adjust curvature-realted bounds with physical limits

* Planning: fix open space pre stop in pull over

* Robot: Remove trailing spaces.

* Robot: Fix typos.

* Robot: Code clean with clang-format.

* update msf localization doc

* add rtk localization tutorial

* change the config for multi-scale yolo obstacle detection (#8693)

* prepare code for multi-scale camera object detection

* add constexpr for const value

* change yolo detecto config for new multi-scale code

* Prediction: implement predict next position in pedestrian model

* Planning: change back dkappa_bound

* Prediction: implement main procedures in pedestrian interaction

* Prediction: small refactor of BUILD in pedestrian interaction model

* Planning: simplfy costheta smoother interface

* Planning: lane alighed pull over heading

* Planning: remove a few AERROR msg

* Planning: param tunings for pull over scenario

* planning: clean up old obseleted files

* Prediction: fix model error

* Prediction: tmp disable pedestrian interaction model

* Prediction: fix pedestrian interaction model core dump error

* planning: open space add naive linear interpolation inside python wrapper to study infeasible problem issue.

* *prediction : fix typo

* Prediction: introduce lane_aggregating_evaluator that uses symmetric functions to aggregate lane_features and classify what lane the obstacle vehicle will select in the future.

* Prediction: extract obstacle features for lane_aggregating_evaluator.

* Prediction: extract all the lane-features for lane_aggregating_evaluator.

* Prediction: encode obstacle features in lane_aggregating_evaluator.

* Prediction: encode all lane features and put them into a vector.

* Prediction: aggregate lane_encodings for lane_aggregating_evaluator.

* Prediction: predict scores for all lane-sequences of lane_aggr_evaluator.

* Prediction: remove useless code.

* cyber: code clean

* Robot: Refactor BUILD files.

* Robot: Remove trailing spaces.

* Robot: Code clean with clang-format.

* Prediction: add stable-softmax function to calculate probabilities.

* commit ncut for lidar segmentation (#8694)

* change usb_cam.cc to supoort both 14.04 and 18.04

* drivers: further fix for camera driver on ubuntu 18.04

* cyber: add build and docker support for Cyber RT development (#8335)

* adding build_cyber option in apollo.sh
* add docker start and into scripts for Cyber RT only docker containers

* cyber: enhancing apollo.sh for Cyber RT development

* adding build_drivers option to build drivers only
* adding options of opt/dbg for build_cyber and build_drivers

enable bazel's feature to compute md5 checksums in parallel.