[all] Set ERL_CRASH_DUMP_SECONDS to a negative value

[mocchira]

In order to generate crash dump files for further digging when the erlang runtime is rebooted by heart 1.
I think the restarts reported on #644 caused by heart and for further digging this case, it would be good to have crash dump files with as much information as possible.

[mocchira]

Note: I decided to keep the current default behavior (not dumping a crash_dump) instead recommend users to uncomment (enable dumping full crash_dump) only if they face leo_(manager|storage|gateway) restarted frequently for further digging.