As a user I want a nicer local installer experience when installing multiple kernels

[mikebroberg]

Expected behavior

Having already used jupyter pixiedust install to install Spark 2.0, I wanted the local installer to suggest a default location for Spark 1.6 that's not /Users/mbrobergus.ibm.com/pixiedust/bin/spark/spark-2.0.2-bin-hadoop2.7. The fix is to suggest /Users/mbrobergus.ibm.com/pixiedust/bin/spark and go from there. The same thing happens with the Scala version.

Actual behavior

local installer suggests /Users/mbrobergus.ibm.com/pixiedust/bin/spark/spark-2.0.2-bin-hadoop2.7 as default directory when I already have 2.0 installed.

Steps to reproduce the behavior

re-run jupyter pixiedust install to install a different kernel