Change infobar creation to use a public static Create() method on the infobar delegate classes.

Change infobar creation to use a public static Create() method on the infobar delegate classes.  Make constructors as private as possible.

This has several purposes:
* By preventing direct instantiation, it prevents callers from leaking if they create an infobar and don't add it to an InfoBarService.
* By moving decision-making about when to show infobars into these Create() functions, there's a pattern for where such code should go, and caller code becomes simpler and easier to read.
* The two bullets above together mean that for infobars which should only be displayed in certain circumstances, code can't accidentally bypass the decision logic.
* It enables us to eliminate a common InfoBarService* temp on the caller side since the caller no longer needs to both pass the pointer to the infobar _and_ call AddInfoBar() on the pointer.  This was also a somewhat redundant-looking pattern.
* It makes it easier to change the ownership model for infobars in the future by limiting the affected callsites to only the Create() functions.

Note that right now, this still feels pretty redundant since we pass all the same args to Create() functions as constructors most times.  In the new ownership model constructors will no longer need to take InfoBarService*s, which will make this better.

Additionally, this makes AddInfoBar()/ReplaceInfoBar() take scoped_ptr<>s to indicate they're receiving ownership.  This sort of change is easy to make since we only need change the create functions.

This change also has a functional effect: it eliminates some cases where we tried to only show infobars when no other infobars were already showing (discussed and approved by Glen).

BUG=none
TEST=none
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=175467

[Peter Kasting, 2012-12-28 02:36:19]

Patchset 9 also includes a change to remove InfoBarDelegate::InfoBarClosed() and
instead have callers just delete objects normally.  I needed this in order to
fix some memory corruption problems with the patch.  If you want I can probably
split this into a precursor change (it will only reduce the change size by like
6 files though).

[Elliot Glaysher, 2013-01-02 20:23:25]

my eyes have rolled over this entire change. nothing jumped out as wrong. lgtm.

[Peter Kasting, 2013-01-02 21:59:39]

Nico, any chance you can rubber-stamp this as a chrome/ OWNER?  I'd send this to
sky but he and Ben both seem to be out.

[Nico, 2013-01-02 23:24:19]

On 2013/01/02 21:59:39, Peter Kasting wrote:
> Nico, any chance you can rubber-stamp this as a chrome/ OWNER?  I'd send this
to
> sky but he and Ben both seem to be out.

ECLTOOBIG

[Peter Kasting, 2013-01-03 19:17:16]

sky: chrome/ OWNERS

[sky, 2013-01-07 17:17:22]

LGTM