Reset the obsolete flag on service thread

[MrZammler]

Summary

This PR makes 2 small changes:

The first is if the service thread does not clean a chart (because of timing factors) to re-set the host wide RRDHOST_FLAG_PENDING_OBSOLETE_CHARTS flag to re-try on the next iteration.

The second change is to execute rrdset_free in all memory modes.

The goal is to see whether memory usage can be reduced in case of many ephemeral charts.

Test Plan

Additional Information

For users: How does this change affect me?

[MrZammler]

This change appears to somewhat alleviate the case of #15066

[MrZammler]

Noticed on an agent running this patch the following during shutdown:

2023-08-31 12:22:26: netdata INFO  : DAEMON_COMMAND : (1696@database/sqlite/sqli:metadata_sync_s): METADATA: Sending a shutdown command
2023-08-31 12:22:26: netdata INFO  : DAEMON_COMMAND : (1701@database/sqlite/sqli:metadata_sync_s): METADATA: Waiting for shutdown ACK
2023-08-31 12:22:26: netdata INFO  : METASYNC : (1675@database/sqlite/sqli:metadata_event_): METADATA: Shutting down event loop
2023-08-31 12:22:26: netdata INFO  : DAEMON_COMMAND : (1704@database/sqlite/sqli:metadata_sync_s): METADATA: Shutdown complete
2023-08-31 12:22:26: netdata INFO  : DAEMON_COMMAND : (0440@daemon/main.c       :netdata_cleanup): NETDATA SHUTDOWN: in       0 ms, stop metasync threads - next: wait for dbengine collectors to finish
2023-08-31 12:22:26: netdata ERROR : DAEMON_COMMAND : (0450@daemon/main.c       :netdata_cleanup): waiting for 2142 collectors to finish (sleeping for 100000 microseconds every time this happens)
2023-08-31 12:22:26: netdata ERROR : DAEMON_COMMAND : (0450@daemon/main.c       :netdata_cleanup): waiting for 2142 collectors to finish (similar messages repeated 3 times in the last 1 secs) (sleeping for 100000 microseconds every time this happens)
2023-08-31 12:22:27: netdata ERROR : DAEMON_COMMAND : (0450@daemon/main.c       :netdata_cleanup): waiting for 2142 collectors to finish (similar messages repeated 10 times in the last 1 secs) (sleeping for 100000 microseconds every time this happens)
....

8 minutes later the same message repeats. Not sure if it's because of this patch or not, will have a look.

[thiagoftsm]

PR worked as expected when I obsolete a thread in eBPF.plugin, LGTM!