xsigo: Hardening driver in handling remote QP failures

xsigo: Hardening driver in handling remote QP failures

Orabug: 24929076

Handle scenario's where PSIF generates batched
transmit completions.

In case of Remote QP disconnect follow this sequence
-> If there are any pending transmit completion
explicitly transition QP state to ERROR.
-> Wait for a maximum of 10 seconds for all pending
completions
(10 seconds derived from retrycount * local ack timeout)
Destroy QP
Wait for another 10 seconds(max) if completions are
not returned by hardware.

Synchronized calls to poll_tx

Added more efficiency in handling of TX queue full condtion.

Handle scenario where uVNIC removal can come in batches

Reported-by: Pradeep Gopanapalli <pradeep.gopanapalli@oracle.com>
Signed-off-by: Pradeep Gopanapalli <pradeep.gopanapalli@oracle.com>
Reviewed-by: Aravind Kini <aravind.kini@oracle.com>
Reviewed-by: viswa krishnamurthy <viswa.krishnamurthy@oracle.com>
Reviewed-by: Manish Kumar Singh <mk.singh@oracle.com>
Reviewed-by: Ariel cohen <ariel.x.cohen@oracle.com>
Reviewed-by: UmaShankar Tumari Mahabalagiri <umashankar.mahabalagiri@oracle.com>