Update parameters for navigation / safe_teleop

[YutoUchimi]

Split from #1143

Summary of commits by @furushchev

• Use 2d costmap: VoxelLayer -> ObstacleLayer.
• Fix laser topics used for obstacle observation in move_base.
• Adjust params for move base tilt observation persistence.
• Update safe teleop params.

Summary of commits by @iory

• Add option to use teb_local_planner.

Summary of commits by @knorth55

• Use move_base local costmap parameter for safe teleop base.

Summary of commits by @YutoUchimi

• Use pentagonal footstep polygon.
• Use omni-corrected model type and set valid parameters for that type.
• Update amcl parameters for robust localization against wheel slipping.
• Limit max_vel_y to avoid running like a crab.
• Decrease sim_time and increase controller_frequency to avoid frequent navigation chattering on real robot which occurs when forward simulation is not finished in the timeout.
• Increase min_obstacle_height so as not to observe some noised points on floor plane from tilt laser.
• Increase padding of tilt_laser_self_filter not to regard PR2's links as obstacle while moving base.
• Explicitly disable recovery_behavior which is hard to notice because max_planning_retries is set to -1 in default, which actually disables recovery_behavior.
• Update params for cost calculation used during global planning.
• Do not allow to plan to go through unknown space.
• Allow small translational movement when rotating in place to enable to move in '-x' direction.

[YutoUchimi]

I chery-picked b58bd64 by mistake.
It should be merged into #1121 , so I reverted that commit from this PR.

[k-okada]

please check on fridge demo with @takayuki5168

[takayuki5168]

@k-okada I checked fridge demo with @HiroIshida 6 times last week. There are some below failures, but anything below has nothing to do with this PR.

• Tilt laser device was not recognized.
  • After rebooting PR2 by circuit breaker, the device was found, but costmap is abnormal. The cost in front of PR2 is light blue, which means PR2 could not move at all.
  • After recalibrating tilt laser twice, the costmap becomes normal, and everything worked with demo. So we think that the calibration of tilt laser became shifted as time passed by.
• TF was not published. After rebooting PR2, everything worked with demo.