Add E-Branchformer module

[TeaPoly]

Hi,
This PR is about the E-Branchformer. Modified based on Branchformer implementation.

The results as follows：

• Feature info: using fbank feature, dither=1.0, cmvn, online speed perturb
• • Model info:
    • Model Params: 47,570,132
    • Num Encoder Layer: 17
    • CNN Kernel Size: 31
• Training info: lr 0.001, weight_decay: 0.000001, batch size 16, 4 gpu, acc_grad 1, 240 epochs
• Decoding info: ctc_weight 0.3, average_num 30
• Git hash: 89962d1

decoding mode	CER
attention decoder	4.73
ctc greedy search	4.77
ctc prefix beam search	4.77
attention rescoring	4.39
LM + attention rescoring	4.22

[xingchensong]

THX, the result looks promising.