why do we consider that all currents flow into the transistor while analyzing the circuit?

To start out, it's NOT the case that "we consider" all currents flowing into or out of a transistor when analyzing a circuit. What IS the case is that we take a consistent view and let the signs work themselves out in the end. Some people, and here I'm referring to Ebers and Moll in their original 1954 paper called "Large-Signal Behavior of Junction Transistors," made the choice to consider all currents as flowing inward. Their Figure 3 shows a PNP set up that way, but their analysis of NPN is the same.

This is also the case with Ian Getreu's "Modeling the Bipolar Transistor" that was originally published by Tektronix circa 1979. Ian is still around, by the way, and I recently helped him republish his book. I have zero financial interest, just felt the book was too important and also unique, so I wanted to see it back in print. It's available at Amazon.

But if you look at the Sedra/Smith link below, you will see two currents inward and one outward (at the emitter) and their equations are set up appropriately, with the sign change for the I_E current, as expected. So it is NOT the case that everyone does it that way. You choose your directions and let the expressions you derive flow out of that and then check the sign of your results to determine if the original "assumption" is correct, or reversed from what was assumed.

Sometimes, it's better to make them all go inward or all go outward, not because it works any better -- it doesn't -- but because it keeps _YOU_ from making hidden assumptions and getting trapped by your own preconceptions. Physicists are very sensitive to this problem. So just to avoid that, it's likely that Ebers and Moll simply chose to pick an impossibility knowing in advance that somehow the signs had better work out in the end -- just as a matter of mental rigor. Sometimes people make assumptions based upon experience about directions and in the end screw things up because they inadvertently mix up their conventions as they write out the equations due to their hidden preconceptions. It's cleaner to just pick a single direction for everything so that you keep your mind in a consistent mode as you write things out.

I've included the Sedra/Smith link and a google link to the original 1954 paper in PDF form below.

You don't have to, but if you want a correct answer, I'd advise it.

Why would you NOT consider all the currents? There are only 3.

edit, rereading, I see that you are actually asking why the convention is that current flows into the transistor, as opposed to out of it.

The answer is, it doesn't matter. You have to pick a direction, so you pick in. You could pick out and get the same results.