[jsk_fetch_startup] Use audible_warning to speak diagnostics

[708yamaguchi]

Duplicate knorth55#223

I use jsk_tools/audible_warning.py instead of jsk_fetch_startup/warning.py's function of speaking diagnostic.
jsk-ros-pkg/jsk_common#1607

I think it is important to unify the function of speaking diagnostic.

[k-okada]

need to fix CI of jsk-ros-pkg/jsk_common#1607

[708yamaguchi]

I force-pushed because

• [jsk_fetch_startup] Use audible_warning to speak diagnostics knorth55/jsk_robot#223 is not currently used.
• Refined version [jsk_fetch_startup] Add audible warning knorth55/jsk_robot#233 is used.

[708yamaguchi]

I think we do not need to wait for upstream release.
The reason is the same as #1481 (comment)

===
In fetch PC, we use knorth55/jsk_common's fetch15 branch.

jsk_robot/jsk_fetch_robot/jsk_fetch.rosinstall.melodic

Lines 31 to 34 in 3d65836

	- git:
	local-name: jsk-ros-pkg/jsk_common
	uri: https://github.com/knorth55/jsk_common.git
	version: fetch15

In this branch, sanity_diagnostics.py is already merged.
knorth55/jsk_common#4

[708yamaguchi]

I force-pushed the commits to resolve conflicts.

[708yamaguchi]

The latest commit f47f29d is cherry-picked from knorth55#282

rosrun rosserial_python seiral_node.py via bluetooth sometimes loses connection.
At that time, serial_node.py generates a diagnostics error as follows.

However, serial_node.py can automatically reconnect after losing connection.
So I think audible_warning should ignore this diagnostics error.

[k-okada]

wait for discussion #1562 (comment)

[sktometometo]

@708yamaguchi Could you please resolve conflicts?