selftests: net: give more time for GRO aggregation

The gro.sh test-case relay on the gro_flush_timeout to ensure
that all the segments belonging to any given batch are properly
aggregated.

The other end, the sender is a user-space program transmitting
each packet with a separate write syscall. A busy host and/or
stracing the sender program can make the relevant segments reach
the GRO engine after the flush timeout triggers.

Give the GRO flush timeout more slack, to avoid sporadic self-tests
failures.

Fixes: 9af771d2

 ("selftests/net: allow GRO coalesce test on veth")
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Tested-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/bffec2beab3a5672dd13ecabe4fad81d2155b367.1706206101.git.pabeni@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>