Fix tests to use password auth as md5

[cpj2195]

No description provided.

[hanefi]

I do not see the value of this PR. Is there any harm in setting POSTGRES_HOST_AUTH_METHOD=trust in a test environment?

[dimitri]

I do not see the value of this PR. Is there any harm in setting POSTGRES_HOST_AUTH_METHOD=trust in a test environment?

I think the idea was that we used to have a problem with parsing connection strings longer than MAXCONNINFO, a static length provided by libpq, size 1024 bytes. We wanted to make sure that with some long authentication entries we would still be able to parse the connection string and connect. Meanwhile I believe we have fixed that issue, but making sure we're good on that front is a good idea.

That said we might be able to only change a couple of the unit test cases (with and without logical decoding) to have regression testing support in that area.