Updated install instructions for new TF and Bazel versions

[zaqqwerty]

Recently did a fresh source build after the update to TF 2.3.0 and Bazel 3.1.0, used what I learned to update the install instructions

[MichaelBroughton]

This looks good. Would you also mind updating the install lines in the Tutorial notebooks ? We shouldnt merge this in until we actually cut a release that incorporates the 2.3.0 upgrade Into the non-nightly/master versions (Otherwise our docs would update prematurely). It will be good to have ready for when we do.

[balopat]

FWIW, this worked for me, thank you, @zaqqwerty, I was running into all sorts of issues with the combination of tools that are in the current version of docs.

[MichaelBroughton]

If we are going to gear up for a 0.4 release in a week or two, would you mind updating this PR to check for any remaining old references ?

Another note is we may want to update the research branch.

[zaqqwerty]

@MichaelBroughton added #395 so we can update the research branch once this is in. Should the workspace file be updated here too? (currently refers to TF 2.2.0)

[MichaelBroughton]

Oops, we just fixed the workspace file, could we also update this to 2.3.1 now in anticipation of our cutting a 0.4 soon ?

[zaqqwerty]

Updated to 2.3.1

[MichaelBroughton]

Looking like we are almost ready to pull the trigger on 0.4.0 TFQ. One thing I noticed going through the notebooks was that

1. PauliString.expectation_from_wavefunction has been deprecated in favor of PauliString.expectation_from_state_vector and TrialResult.final_state has been deprecated in favor of TrialResult.final_state_vector. The calculate gradients tutorial and the hello many worlds tutorial both raise warnings on this. Would you mind updating this in the PR to move to the new names ?

I've checked all the tutorials and they seem to function on the latest tfq-nightly so I think after we fix these deprecation warnings we can merge this PR and #395 Either Tomorrow or the day after and then cut the release 🎉

[zaqqwerty]

Addressing the Cirq deprecations in #403