Perf: avoid concat on flush as well

[Beanow]

This only helps when we're not zero padding the last buffered data. Otherwise we'll always have at least 2 buffers and concat.

singleRemainder: test({
	inSize: 150*MB,
	inCount: 1,
	blockSize: 160*MB,
	zeroPadding: false
}),
bigRemainder: test({
	inSize: 200*MB,
	inCount: 1,
	blockSize: 101*MB,
	zeroPadding: false
})

Test	Version	AVG
singleRemainder	pull-block master	68ms
singleRemainder	pull-block dev	0ms
bigRemainder	pull-block master	56ms
bigRemainder	pull-block dev	0ms

[Beanow]

Full blown chromium results

Test	Version	Min	AVG	Max
singleRemainder	master	46ms	59ms	86ms
singleRemainder	dev	0ms	0ms	0ms
bigRemainder	master	30ms	38ms	70ms
bigRemainder	dev	0ms	0ms	0ms
perfectHalves	master	0ms	0ms	1ms
perfectHalves	dev	0ms	0ms	1ms
almostHalves	master	18ms	22ms	30ms
almostHalves	dev	18ms	23ms	53ms
perfect10ths	master	0ms	1ms	2ms
perfect10ths	dev	0ms	1ms	2ms
almost10ths	master	4ms	5ms	8ms
almost10ths	dev	4ms	5ms	8ms
perfect100ths	master	1ms	2ms	3ms
perfect100ths	dev	1ms	2ms	4ms
almost100ths	master	1ms	3ms	6ms
almost100ths	dev	1ms	2ms	3ms
hugeBuffers	master	1ms	1ms	2ms
hugeBuffers	dev	1ms	1ms	2ms
tinyBufs	master	43ms	50ms	81ms
tinyBufs	dev	40ms	50ms	75ms
microBufs	master	51ms	52ms	64ms
microBufs	dev	51ms	59ms	88ms
manyPairs	master	41ms	58ms	87ms
manyPairs	dev	43ms	71ms	79ms
perfectPairs	master	41ms	64ms	76ms
perfectPairs	dev	39ms	60ms	75ms
manyTriplets	master	44ms	72ms	79ms
manyTriplets	dev	44ms	62ms	79ms
bigBuffers	master	1ms	1ms	2ms
bigBuffers	dev	0ms	1ms	4ms

[Beanow]

Fun fact, the original PR #2 was also due to concat, not slice. As a concat of 1 buffer copies the buffer to a new one. This happened a lot before 405d1ab

[dignifiedquire]

Thank you 🎉