Added telegram_print function

[apthagowda97]

Thanks for this wonderful library.

When I commit a kernel in Kaggle I use knockkncok(telegram_sender) so that I 'll get to know the status of training. But the problem is I am unable to view some important data such as loss or accuracy for each epoch while the model is training.

So, I thought of writing a simple function

telegram_print(text="<your_text>",token="<your_api_token>", chat_id=CHAT_ID)

which can be used to send the message directly to telegram while training. It helps a lot to kill the kernel when the model is not performing well and thereby saving GPU.