More resolv.conf features

[brzuchal]

This PR resolves #47 All added features works the same way as glibc implementation of resolv.conf Changes:

• added search setting support
• added domain setting support
• added options ndots:X support, where X is positive integer, silently capped to 15
• added options rotate support, which balance using round robin between nameservers
• fixed options timeout X now works with with : delimiter in format options timeout:X
• fixed options attempts X now works with with : delimiter in format options attempts:X
• env LOCALDOMAIN which overrides search and domain
• env RES_OPTIONS which overrides default options like ndots, timeout, attempts, rotate
• when no search list provided and no env LOCALDOMAIN given gethostname() is used

Valid resolv.conf format (from Travis CI):

# Dynamic resolv.conf(5) file for glibc resolver(3) generated by resolvconf(8)
#     DO NOT EDIT THIS FILE BY HAND -- YOUR CHANGES WILL BE OVERWRITTEN
nameserver 169.254.169.254
search c.travis-ci-prod-2.internal google.internal
options timeout:1
options attempts:3
options ndots:3
options rotate

All max and default values are taken from resolv/res_init.c and from man

[brzuchal]

Tests not passing on Travis using 169.254.169.254 nameserver. The issue was in options timeout:1 cause its unit is seconds, not milliseconds.

[brzuchal]

Integration test fails on x64 platform on AppVeyor randomly.

1) Amp\Dns\Test\IntegrationTest::testResolve with data set #2 ('stackoverflow.com')
Amp\Dns\DnsException: All query attempts failed for stackoverflow.com: No response for 'stackoverflow.com' (A) from any nameserver after 2 attempts, tried udp://8.8.8.8:53, udp://8.8.4.4:53, No response for 'stackoverflow.com' (AAAA) from any nameserver after 2 attempts, tried udp://8.8.8.8:53, udp://8.8.4.4:53

@kelunik shouldn't we use default OS settings there?

[kelunik]

@brzuchal Probably, yes.

Merged, thanks! :tada: