How to estimate the first user story when you start the first sprint?

You could try doing a team sort to begin with, initially using T-Shirt sizes, and determine numbers afterwards:

https://dzone.com/articles/agile-estimation-practice

Should I take a user story and give it a number (in Fibonacci series) and ask the team to estimate the other stories? Or is there any better way to do so?

Think about who will work on this first story ? and if its the team then who should ideally give efforts ? How about let team decide together and give it an estimate. You can start with story points ( fibonacci) or any other method and see how it works for your team.

Should I take a user story and give it a number (in Fibonacci series) and ask the team to estimate the other stories? 

Why would you give a story a number?  Are you going to be doing any of the work?  

Both of the above responses are good.  I agree with @Ian about the t-shirt sizing to start because that is easier for people to relate. What ever way to choose to go, as Scrum Master you should not be doing any of the estimation.  As the Scrum Guide says 

The Development Team is responsible for all estimates. 

It is your job as Scrum Master to help them understand how to do relative estimation and coach them on improvement.  They will be bad at it in the beginning.  Help them get better at it as they go.

I noticed something I want to address:

How do I as an agile coach/scrum master tell team to estimate their first user story?

You ask how you tell the team to estimate. My question would be: why would you want to? A better question to ask may be something like 'how do I help the team discover good ways to estimate?'

Then for the actual question. Firrstly, in my opinion specific story estimates in the first few sprints for stories related to a new product are not trustworthy anyway, so worrying about getting those initial estimates right is a waste of time and effort. What is much more important is that you take every opportunity you get to review the sizing of the stories as the team progresses through them, and what you can learn in order to hone upcoming estimates and evolve a good understanding of what the stories entail for the product in question and how to refine them. It's a great discussion topic for early retrospectives.

Also I would say it depends heavily on the maturity of the team and the familiarity of the work involved. If it's a very experienced and stable senior team working on their umpteenth product Scrum-style for the same company, they can probably just pretty much go by their gut and quickly evolve a good system for estimating in the first few sprints. But if you're talking about a team that's new to development and/or agile ways of working, some more coaching/teaching work may be in order. For example, I am personally quite enamored of the paint the room and ball point game training exercises to get people thinking about it :)

Thanks everyone for your responses.

As we do an inspect and adapt in scrum, good estimate evolve during the course of the sprints. Hence it is a good idea to start with T-Shirt sizing and then try to relate a number to it during the initial sprint and then once the team matures, they would know what to do.

One of the methods that I have heard scrum masters follow initially is to not take any user story more than 13 story points in a sprint. If a story is more than 13, then work on splitting the story to smaller stories. The number (13) may vary from team to team.