Remove buffer to decrease copy when encode NSString

[zhongwuzw]

m_formatBuffer is not needed, leads to additional malloc and times of copy.

[lingol]

Have you run the baseline test before & after this modification? Is there any difference for NSString encoding?

Logically speaking, it will be faster after this optimization. It's better to know.

[lingol]

Keep indentation aligned please.

[zhongwuzw]

@lingol Yes, it's more faster, but TBO, using baseline test result to compare is unfair, because this optimize only have benefits about 10ms under 10000 times, every time I run test, it's hard to figure out wether it works, because the total time of baseline test ranges from 300ms to 400ms.

So I used instrument to profile, but you know Time Profiler just do samples, so just for reference. From the image below, we can see the consume of old way. It capture malloc、getBytes:maxLength:usedLength:encoding:options:range:remainingRange:、+ dataWithBytes:length:、- maximumLengthOfBytesUsingEncoding: etc.

The new way, Time Profiler capture the consume of dataUsingEncoding, about 9ms,