cashew: reset local counter on plan transition

cashew: reset local counter on plan transition

The local counter fails to estimate data usage correctly if it is active
during a transition from one plan to another. This CL is the first of
two that will resolve this issue. It addresses plan transitions due to
data quota exhaustion. The subsequent CL will address plan transitions
due to the expiration time having passed.

Changes:
- Add the ability to reset the local byte counter.
- Update OnByteCounterUpdate event handler to iterate over all active
  plans and assign measured data traffic appropriately to each plan in
  turn. If a plan is consumed, any overage is applied to the next active
  plan and the local byte counter is reset accordingly. This continues
  until we're either out of data to assign or out of active plans.
- Move the bulk of the ServiceImpl::OnByteCounterUpdate logic to the
  DataPlan class to isolate it for unit testing.

BUG=chromium-os:9860
TEST=unit tests, manual testing on device

Change-Id: Ibfe2ee15eb6b79cb4f6575f289629be8388e12f7

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=738b13e

[Jason Glasgow, 2011-02-08 14:56:46]

LGTM

[Eric Shienbrood, 2011-02-10 18:47:00]

LGTM, with a couple naming suggestions that you're free to ignore if you like.
Also, I didn't read every line of each new unit test, but I scanned through them
and check a few in detail.

http://codereview.chromium.org/6250171/diff/1/src/data_plan.h
File src/data_plan.h (right):

http://codereview.chromium.org/6250171/diff/1/src/data_plan.h#newcode30
src/data_plan.h:30: typedef int64 Bytes;
Would "ByteCount" be a better name for this? "Bytes" sounds like something you'd
use to refer to an array of bytes.

http://codereview.chromium.org/6250171/diff/1/src/device.h
File src/device.h (right):

http://codereview.chromium.org/6250171/diff/1/src/device.h#newcode70
src/device.h:70: // |parent_| must already have received at least one update
parent_ is in the implementation class, so it's kind of weird to refer to it
here. Probably better to say "Parent service must already have received one
update"

[Vince Laviano, 2011-02-10 22:29:19]

http://codereview.chromium.org/6250171/diff/1/src/data_plan.h
File src/data_plan.h (right):

http://codereview.chromium.org/6250171/diff/1/src/data_plan.h#newcode30
src/data_plan.h:30: typedef int64 Bytes;
On 2011/02/10 18:47:00, Eric Shienbrood wrote:
> Would "ByteCount" be a better name for this? "Bytes" sounds like something
you'd
> use to refer to an array of bytes.

Done.

http://codereview.chromium.org/6250171/diff/1/src/device.h
File src/device.h (right):

http://codereview.chromium.org/6250171/diff/1/src/device.h#newcode70
src/device.h:70: // |parent_| must already have received at least one update
On 2011/02/10 18:47:00, Eric Shienbrood wrote:
> parent_ is in the implementation class, so it's kind of weird to refer to it
> here. Probably better to say "Parent service must already have received one
> update"

Done.

[Eric Shienbrood, 2011-02-10 22:33:23]

LEBTM (looks even better to me)