Add a new retrydns NFS mount option

[rmacklem]

Summary

Without this patch, an NFS mount attempt will
fail if the NFS server name cannot be resolved via
getaddrinfo(). This failure may be caused for a
mount specified in fstab(5) if the DNS service is
slow to startup.

This patch adds a new mount option called "retrydns=<N>"
which allows for retries of the getaddrinfo() call.
The first retry is done after 1sec and the delay is backed
off to 16sec.

A setting of "retrycnt=0" retries until success.

A mount_nfs.8 man page patch will be done in a separate review.

Test Plan

Tested for trivial cases, such as a mistyped server name.
I am hoping glebius@ can test it for his failure case.

[glebius]

I tested the patch and the patch works as expected. With retrydns=10 I see in console.log:

Feb 25 22:48:33 shop kernel: Mounting NFS filesystems:mount_nfs: myhost.org:2049: Name does not resolve
Feb 25 22:48:33 shop kernel: mount_nfs: myhost.org:2049: Name does not resolve
Feb 25 22:48:33 shop syslogd: last message repeated 2 times
Feb 25 22:48:33 shop kernel: .

However, I don't agree on need for new option and solution that doesn't involve the backgrounding loop. I will reply in the email thread. Looking at the most recent replies from Cy and Chris, it seems that I was not clear that I want modify behavior only of background mounts, not affecting blocking mounts at all. I'll followup.