improve the perf

For some big swagger, it seems too slow, the problem is that the tool does the below thing during its run
1 use autorest(v2) to transform the swagger and get a resolved swaggger and source mapping file
2 transform the swagger allOf and path level parameter to make it fit the modeling which is defined in C#

when we consider improving the perf, we can focus on above points

Start by profiling the two stages named in the issue: AutoRest v2 resolution with source mapping, and the transformations for Swagger allOf and path-level parameters. Compare runs on a large Swagger document; done should mean measurable performance improvement, but the issue does not specify files, tests, or a target benchmark.

索引モデルが issue の本文から書いたものです。