feat(fs): let a file crawl leave symbolic links alone

[marevol]

FileSystemClient walks a directory with java.io.File, which follows a symbolic
link and offers no way not to. A link is therefore crawled as though it were the
directory or the file it points at, and the target is indexed a second time --
under the crawling configuration that reached the link, not the one that covers
the target.

The second copy carries the permissions of the link. Where a share is readable
by more people than the tree it links into, anyone who can create a link on that
share can have documents from the other tree re-indexed under their own roles
and read them back through search. The same file also ends up in the index
twice, which distorts the hit count and the ranking.

Until now the only way to avoid it was to name the link in the excluded paths of
every configuration. A followSymlink init parameter is added instead. It
defaults to true, so nothing changes for an existing crawl; set
client.followSymlink=false in the configuration parameters of a file crawling
configuration and a link is refused with 403 and left out of the child urls of
the directory it sits in, with one INFO line naming what was skipped.