Now, since my programming skill is very poor, this `PrefetchedDataLoader` only allow generate a single iter at the same time.
the benefit of `PrefetchedDataLoader` is that, `PrefetchedDataLoader` provides better performance with a simple replacement in most of the existing codes.
test:
```python
$ cat iternew.py && python iternew.py
import mxnet as mx
from mxnet.gluon.data import PrefetchedDataLoader as DataLoader,ArrayDataset
from time import sleep,perf_counter_ns
train_data=ArrayDataset(mx.nd.array([[i] for i in range(50000)]),mx.nd.array([[99-i] for i in range(50000)]))
test_data=ArrayDataset(mx.nd.array([[i] for i in range(10000)]),mx.nd.array([[99-i] for i in range(10000)]))
def transform_train(sample):
  sleep(0.0016)
  return sample

def transform_test(sample):
  sleep(0.0008)
  return sample

train_iter=DataLoader(train_data.transform_first(transform_train),batch_size=500,num_workers=10)
test_iter =DataLoader(test_data .transform_first(transform_test ),batch_size=500,num_workers=10)
if True:
  tic=perf_counter_ns()
  for epoch in range(10):
    print("epoch"+str(epoch)+" start at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
    for i in train_iter:
      sleep(0.1)
    print("       finished train phase at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
    for i in test_iter:
      sleep(0.05)
    print("        finished test phase at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
  print("cost="+str((perf_counter_ns()-tic)*1e-9)+"s")

epoch0 start at 0.0s
       finished train phase at 11.25s
        finished test phase at 12.31s
epoch1 start at 12.31s
       finished train phase at 22.62s
        finished test phase at 23.68s
epoch2 start at 23.68s
       finished train phase at 34.03s
        finished test phase at 35.09s
epoch3 start at 35.09s
       finished train phase at 45.41s
        finished test phase at 46.48s
epoch4 start at 46.48s
       finished train phase at 56.82s
        finished test phase at 57.88s
epoch5 start at 57.88s
       finished train phase at 68.24s
        finished test phase at 69.3s
epoch6 start at 69.3s
       finished train phase at 79.65s
        finished test phase at 80.71s
epoch7 start at 80.71s
       finished train phase at 91.04s
        finished test phase at 92.11s
epoch8 start at 92.11s
       finished train phase at 102.46s
        finished test phase at 103.53s
epoch9 start at 103.53s
       finished train phase at 113.89s
        finished test phase at 114.95s
cost=114.94954171600001s
```
(cost ~`129.67192333600002s` if we are using `Dataloader` rather than `PrefetchedDataLoader`)

there already exists some faster dataloader in mxnet 2.0, but in v1.x, the exist dataloader is slower and could be improved by changing its prefetch behavior as what 2.0 have done.
```python
$ cat iternew.py && python iternew.py
import mxnet as mx
from mxnet.gluon.data import PrefetchedDataLoader as DataLoader,ArrayDataset
from time import sleep,perf_counter_ns
train_data=ArrayDataset(mx.nd.array([[i] for i in range(50000)]),mx.nd.array([[99-i] for i in range(50000)]))
test_data=ArrayDataset(mx.nd.array([[i] for i in range(10000)]),mx.nd.array([[99-i] for i in range(10000)]))
def transform_train(sample):
  sleep(0.0016)
  return sample

def transform_test(sample):
  sleep(0.0008)
  return sample

train_iter=DataLoader(train_data.transform_first(transform_train),batch_size=500,num_workers=10)
test_iter =DataLoader(test_data .transform_first(transform_test ),batch_size=500,num_workers=10)
if True:
  tic=perf_counter_ns()
  for epoch in range(10):
    print("epoch"+str(epoch)+" start at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
    for i in train_iter:
      sleep(0.1)
    print("       finished train phase at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
    for i in test_iter:
      sleep(0.05)
    print("        finished test phase at "+str(round((perf_counter_ns()-tic)*1e-9,2))+"s")
  print("cost="+str((perf_counter_ns()-tic)*1e-9)+"s")
epoch0 start at 0.0s
       finished train phase at 11.28s
        finished test phase at 12.35s
epoch1 start at 12.35s
       finished train phase at 22.73s
        finished test phase at 23.79s
epoch2 start at 23.79s
       finished train phase at 34.15s
        finished test phase at 35.21s
epoch3 start at 35.22s
       finished train phase at 45.59s
        finished test phase at 46.66s
epoch4 start at 46.66s
       finished train phase at 57.01s
        finished test phase at 58.07s
epoch5 start at 58.07s
       finished train phase at 68.43s
        finished test phase at 69.5s
epoch6 start at 69.5s
       finished train phase at 79.87s
        finished test phase at 80.93s
epoch7 start at 80.93s
       finished train phase at 91.3s
        finished test phase at 92.37s
epoch8 start at 92.37s
       finished train phase at 102.74s
        finished test phase at 103.8s
epoch9 start at 103.8s
       finished train phase at 114.17s
        finished test phase at 115.23s
cost=115.23376344s
```

add unittest for PrefetchedDataLoader

update document

previous test shows that there may be something wrong with the `_MultiWorkerIter` according to the inappropriate __iter__() is called, I tried to fix it by moving the call here.

the added flag is set to `True` rather than the default `False` since in mxnet 2.0, the default `nopython` mode prefetch data and auto reload it.

merge `prefetcheddataloader` into `dataloader`

remove whitespace

solve the warning in L726

fix  typo

fix the outdated perfetcheddataloader

This reverts commit 2c8d858.