Reland of land "Change fallback font collection in HarfBuzzShaper"

Reland of land "Change fallback font collection in HarfBuzzShaper" (patchset #1 id:1 of https://codereview.chromium.org/1235163007/)

Reason for revert:
The breakage here was a side-effect of the change that caused the problem [1].  Rolling this back in therefore.

[1] https://codereview.chromium.org/1234183003



Original issue's description:
> Revert of Reland "Change fallback font collection in HarfBuzzShaper" (patchset #1 id:1 of https://codereview.chromium.org/1241613006/)
> 
> Reason for revert:
> Broke Linux MSAN bots.
> 
> http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Linux%20MSAN/builds/4413
> 
> See stdio log of 4412 build: many reports use on an uninitialized variable in shaping font fallback, about 100+ tests failing that way.
> 
> Original issue's description:
> > Reland "Change fallback font collection in HarfBuzzShaper"
> > 
> > Change HarfBuzzShaper to not populate the fallback font list during text
> > shaping and instead populate it based on the ShapeResult objects at call
> > time. This circumvents additional registration work and renders the text
> > shaping API simpler and more consistent.
> > 
> > Originally landed in r198813 (bbb929c88fbe43979dbb9b2b20218538) and then
> > reverted in r198818 (64e30826c261ac7a2607e891dbe7e9d2d251e5f3) by virtue
> > of Mac 10.6 test failures caused by a missing NULL check, since added.
> > 
> > This is a speculative fix.
> > 
> > R=leviw@chromium.org
> > 
> > Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=198886
> 
> TBR=leviw@chromium.org,eae@chromium.org
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
> 
> Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=198923

TBR=leviw@chromium.org,eae@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=198951

[Noel Gordon, 2015-07-15 12:14:59]

Created Reland of land "Change fallback font collection in HarfBuzzShaper"

[commit-bot: I haz the power, 2015-07-15 12:15:02]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1229413005/1

[commit-bot: I haz the power, 2015-07-15 12:15:25]

Committed patchset #1 (id:1) as
https://src.chromium.org/viewvc/blink?view=rev&revision=198951

[Noel Gordon, 2015-07-15 15:09:07]

So MSAN bots were green after reverting
https://codereview.chromium.org/1234183003

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Linux%20MSAN/bu...
... 4447

Then this change relanded your patch, and the MSAN bots went red, sorry to say. 
So I will need to revert, apologies for the noise.

[Noel Gordon, 2015-07-15 15:11:46]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/1237283004/ by noel@chromium.org.

The reason for reverting is: Broke Linux MSAN bots on relanding:

http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Linux%20MSAN/bu...

==4==WARNING: MemorySanitizer: use-of-uninitialized-value
07:37:45.901 15833       #0 0x7fc50442fdbf in
blink::ShapeResult::fallbackFonts(WTF::HashSet<blink::SimpleFontData const

seems to be the common complaint in the tests that broke..