Replaced access path bases with access paths in backward dataflow

[MForest7]

Replaced nested objects in backward dataflow with access paths, also get a performance increase.

project	stmts	min time	max time	avg time	median time	%
project1	9914	1.4s	1.8s	1.6s	1.5s	39.4
project2	26679	1.3s	1.5s	1.4s	1.4s	16.5
project3	4272	0.3s	0.4s	0.4s	0.4s	28.8
project4	4584	0.3s	0.4s	0.3s	0.3s	35.6
project5	121091	8.9s	9.8s	9.3s	9.0s	37.4
project6	28548	1.6s	1.8s	1.7s	1.7s	38.9
project7	5162	0.4s	0.4s	0.4s	0.4s	28.0

[github-advanced-security]

detekt found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.

[Lipen]

What does the % mean in the table? Replace the table with fewer columns: project, size (stmts or similar), old time (avg +- sd), new time (avg +- sd). If the deviation is low (and it should be low, since the analysis should be near-deterministic), it can be omitted.

What about the "quality"? Comparing only the performance is not enough for such major change. Pick a metric (number of inferred types, ratio of all type, ratio of annotated type, number of inferred calls, etc) and add "old/new quality" columns.