Add criterion benchmarks for common uses of library (signing a transaction, using TransactionBuilder, etc). Also add criterion benchmarks for some interpreter methods (mostly Coll methods since they're generally the most expensive).

For this issue add the following benchmarks:

• Signing a transaction (for now each input box should just be guarded by public key), benchmark should measure scaling with inputs (in case tx Validation is merged, Transaction Validation #736 some validation benchmarks would be nice to have as well), (ergo-lib)
• Coll ops: Map, SizeOf, CalcSha256, Slice, flatMap, etc (these benchmarks should be added to ergotree-interpreter)