Investigate possibility of not modifying instance variables for resetting sequence names

[rpbaltazar]

reset_sequence_names method implementation for postgresql adapter removes the variable sequence_name if it exists. this has the following note:

# sequence_name contains the schema, so it must be reset after switch
# There is `reset_sequence_name`, but that method actually goes to the database
# to find out the new name. Therefore, we do this hack to only unset the name,
# and it will be dynamically found the next time it is needed

but as the name indicates it, it feels hackish and seems to have caused some random issues as described in #81

This logic should be revisited and improved if possible