pull request: fail on form errors - don't redirect

The error page might not be pretty but it is very usable - the user can go back and fix the problem.

pull requests: wrap long lines in diff instead of truncating them

A horizontal scrollbar would not be a convenient solution.

indexer: rework get_node to handle alternative path element separators

Problem seen on Windows by Dominik Ruf.

pull requests: use formencode.Set for initial reviewer list

This works with formencode 1.2.6 too.

This also makes it possible to have an empty initial reviewer list.

pull requests: make it possible to add comments to closed PRs

You can close a PR but you cannot close the discussion of it!

mention: improve pattern for matching multiple mentions

It would (silently as always) fail to recognize some @mentions - especially
when only separated by a single whitespace char.

vcs: replace recursive implementation of prev&next with iterative

It was recursing over the repo and could give "maximum recursion depth
exceeded".

The implementation is still inefficient.

db_manage: don't ask for confirmation before using symlink - just warn

Mostly a backout of bbaf0b86a1fe ... but we still use the normalized path.

Escape shell command parts for Git backend

Python 2.x uses pipes.quote and Python 3.3+ uses shlex.quote in
GitRepository._run_git_command()

pull requests: fix same git repository pull-request issue

Currentry, Kallithea can not create pull-request from same git repository.
Because, ancestor variable is not initialized when create pull-request from same git repository.

This patch find ancestor by `git merge-base` command and initialize it.

remove unused javascript

.diff-menu-activate is removed in 929fc8d926d5

javascript: replace YUE.onDOMReady with $(document).ready

Different execution order of YUE.onDOMReady and $(document).ready makes it hard to do gradually.

notifications: insert 'References' mail headers to help MUA threading

gmail is however fubar.

helpers: user email can be unset

Check for not being empty, not for ''.
Also, don't call person() with email, as it can't be caught properly.

compare: introduce .cs_repo as the repo for .cs_changes - sometimes it is org, sometimes other

Influences pull request creation, display and compare.

pull requests: allow closing of PRs in all states - undo #677

There can be reasons to close a PR without approving or rejecting it -
especially when approving or rejecting a PR overwrites the "review status" of
all changesets in it.

If some code for some reason is interested in PRs that has been closed as
approved or rejected then that is what they should look for ... even though
that is very hard with the current data model.