Fix regression where font size too small for IME short name

ash/common/system/chromeos/ime_menu/ime_list_view.cc

Index: ash/common/system/chromeos/ime_menu/ime_list_view.cc
diff --git a/ash/common/system/chromeos/ime_menu/ime_list_view.cc b/ash/common/system/chromeos/ime_menu/ime_list_view.cc
index e858d695a5210097218fdd30d02054f7c7cdb70c..e3303da79416f6fba95b100b298ba4a97ec2f7f5 100644
--- a/ash/common/system/chromeos/ime_menu/ime_list_view.cc
+++ b/ash/common/system/chromeos/ime_menu/ime_list_view.cc
@@ -105,7 +105,8 @@ class ImeListItemView : public ActionableView {
     // |kMenuIconSize| is not enough. The label will trigger eliding as "I..."
     // or "...". So we shrink the font size until it fits within the bounds.
     int size_delta = -1;
-    while (id_label->GetPreferredSize().width() > kMenuIconSize &&
+    while ((id_label->GetPreferredSize().width() -
+            id_label->GetInsets().width()) > kMenuIconSize &&

[tdanderson, 2017/02/10 16:16:45]

Can you please post a screenshot to the bug of what various IME short names look
like with this CL applied (be sure to include INTL, EXTD, and the ones that are
present in the screenshots attached to #0 on the bug).

I'm confused about what caused this to regress between m-57 and m-58, do you
have any ideas?

[Azure Wei, 2017/02/13 01:07:29]

Sure, attached screenshot to crbug.com/690563. Please take a look.

I've tried some recent CLs on this file, but failed to find the one causing the
regression. I suspected it may caused by some CL related to theme. 

But anyway, I think we could use Labels::GetTextSize() to avoid such regressing.
Since GetTextSize() is protected in Labels, I used
GetPreferredSize()-GetInsets() to get it.

[tdanderson, 2017/02/13 17:03:49]

Thanks a lot, and sgabriel@ has signed off on this.
Acknowledged.

            size_delta >= kMinFontSizeDelta) {
       id_label->SetFontList(base_font_list.DeriveWithSizeDelta(size_delta));
       --size_delta;