Fix IceGrid Glacier2 session managers to keep sessions alive

[pepone]

This PR fix the Glacier2 session managers implemented in IceGrid to correctly keep sessions alive.

[bernardnormier]

Looks good.

[bernardnormier]

I would add virtual for consistency. "virtual" stands for override.

[bentoi]

It's not clear why this is needed. Keep alive was deprecated in favor of ACM heartbeats and afaict this PR has nothing to do with Glacier2 session managers.

[pepone]

It's not clear why this is needed

IceGrid implements two Glacier2 session managers documented in https://doc.zeroc.com/ice/latest/ice-services/icegrid/glacier2-integration-with-icegrid#Glacier2IntegrationwithIceGrid-ResourceAllocationusingGlacier2andIceGrid

Glacier2 router, send ice_ping messages to sessions to ensure they are keep alive, this is done from the Heartbeat callback

ice/cpp/src/Glacier2/SessionRouterI.cpp

Line 431 in 5efb9f4

_sessionRouter->refreshSession(connection);

ice/cpp/src/Glacier2/SessionRouterI.cpp

Lines 908 to 912 in 5efb9f4

	//
	// Ping the session to ensure it does not timeout.
	//
	session->begin_ice_ping(Ice::newCallback(
	new PingCallback(this, callback, current.con), &PingCallback::finished));

This doesn't work for IceGrid/SessionManager or IceGrid/AdminSSLSessionManager which are only keep alive by calling keepAlive.

Glacier2 sessions are added to IceGrid reaper without a connection, which means there is no ACM callback added to them.

ice/cpp/src/IceGrid/SessionI.cpp

Line 340 in 5efb9f4

_reaper->add(new SessionReapable<SessionI>(_database->getTraceLevels()->logger, session), timeout);