Accelerate point cloud conversion to ROS messsge

[zygfrydw]

I this merge request I have significantly accelerated point cloud conversion to ROS message.
I have noticed that create_cloud function is called with NumPy array representing points, therefore the points are already stored in a byte array in the same format as the message format.
According to documentation, the points argument can be any list of iterables, therefore I have added the helper function _check_if_fast_byte_coversion_avaliable, which checks if the fast path can be used in this case.
Most likely the create_cloud is only called with np.ndarray for points, but I am not familiar with this codebase to ensure it.

Benchmark results:

ROS Hz results before changes:

# ros2 topic hz /carla/ego_vehicle/vehicle_status -w 20
average rate: 10.965
        min: 0.086s max: 0.098s std dev: 0.00271s window: 20
average rate: 10.866
        min: 0.087s max: 0.103s std dev: 0.00326s window: 20

Benchmark of carla-bridge with py-spy

# sudo ~/.local/bin/py-spy top --pid $(pgrep bridge) --native
Total Samples 400
GIL: 69.00%, Active: 116.00%, Threads: 10

  %Own   %Total  OwnTime  TotalTime  Function (filename:line)                                                                        
 20.00%  20.00%   0.770s    0.770s   sched_yield (libc-2.31.so)
 15.00%  15.00%   0.530s    0.530s   pthread_cond_timedwait@@GLIBC_2.3.2 (libpthread-2.31.so)
  5.00%  22.00%   0.260s    0.940s   create_cloud (carla_ros_bridge/sensor.py:328)
  5.00%   5.00%   0.150s    0.150s   do_futex_wait.constprop.0 (libpthread-2.31.so)
  3.00%   3.00%   0.070s    0.070s   0x7f2941ef2a8f (numpy/core/_multiarray_umath.cpython-38-x86_64-linux-gnu.so)
  1.00%   3.00%   0.070s    0.170s   can_execute (rclpy/executors.py:446)

ROS HZ after my modifications:

# ros2 topic hz /carla/ego_vehicle/vehicle_status -w 20
average rate: 16.112
        min: 0.056s max: 0.071s std dev: 0.00347s window: 20
average rate: 16.529
        min: 0.057s max: 0.067s std dev: 0.00265s window: 20

Benchamark with py-spy after modifications

# sudo ~/.local/bin/py-spy top --pid $(pgrep bridge) --native
Total Samples 400
GIL: 69.00%, Active: 116.00%, Threads: 10

  %Own   %Total  OwnTime  TotalTime  Function (filename:line)                                                                        
 34.00%  34.00%   13.78s    13.78s   sched_yield (libc-2.31.so)
 14.00%  14.00%    8.52s     8.54s   pthread_cond_timedwait@@GLIBC_2.3.2 (libpthread-2.31.so)
  2.00%   2.00%    1.68s     1.68s   do_futex_wait.constprop.0 (libpthread-2.31.so)
  0.00%   0.00%    1.07s     1.07s   0x7f26641086c3 (libc-2.31.so)
  2.00%   2.00%   0.890s    0.890s   syscall (libc-2.31.so)
  2.00%   2.00%   0.870s    0.900s   __add__ (rclpy/waitable.py:42)
  0.00%   0.00%   0.700s    0.790s   __add__ (rclpy/waitable.py:40)

---

This change is 

[AzaelCicero]

@joel-mb Who else can review this change in order to get it merged?