Assumed values for hashsize and nf_conntrack_max

[donbowman]

This is part of the theme of assuming we know better than how the host is configured.

the code writes the value of 131072 to /sys/module/nf_conntrack/parameters/hashsize.
It then checks that hashsize is >= nf_conntrack_max/4

But it didn't set nf_conntrack_max. So it then fails saying:
Failed to start cluster: hashsize(131072) should be greater than nf_conntrack_max/4 (589824).

I don't think it is correct to be changing these system global variables, it interferes with other sw installed on the host.

[donbowman]

This was referenced from #14
but the 'fix' assumes 131072 is the right value where the auto-tuner in kube-proxy is assuming 589824 for the memory size of my system.