eval: performance improvements

[pgavlin]

These changes improve evaluation performance by memoizing exported values and merged object keys and avoiding copies when building object schemas.

This gives a tremendous improvement in execution time for evaluation-dominated scenarios, but results in little to no change for open- or load-dominated scenarios. However, the memoization and avoidance of copies cuts memory usage in half in all cases.

Strictly speaking, this is a breaking change for consumers of the schema package, but I think that it is worth taking given the improvements.

Local benchmark results:

goos: darwin
goarch: arm64
pkg: github.com/pulumi/esc/eval
cpu: Apple M1 Max
BenchmarkEval-10           	    453	  2566938 ns/op	2204129 B/op	  18883 allocs/op
BenchmarkEval-10           	    468	  2586703 ns/op	2204146 B/op	  18884 allocs/op
BenchmarkEval-10           	    462	  2588916 ns/op	2204202 B/op	  18883 allocs/op
BenchmarkEval-10           	    464	  2586365 ns/op	2204228 B/op	  18884 allocs/op
BenchmarkEval-10           	    463	  2577383 ns/op	2204362 B/op	  18884 allocs/op
BenchmarkEval-10           	    477	  2537640 ns/op	2204389 B/op	  18884 allocs/op
BenchmarkEval-10           	    468	  2582930 ns/op	2204594 B/op	  18884 allocs/op
BenchmarkEval-10           	    463	  2582915 ns/op	2204246 B/op	  18884 allocs/op
BenchmarkEval-10           	    469	  2608014 ns/op	2204382 B/op	  18884 allocs/op
BenchmarkEval-10           	    465	  2554270 ns/op	2204313 B/op	  18884 allocs/op
BenchmarkEvalOpen-10       	      9	119163125 ns/op	2208651 B/op	  18926 allocs/op
BenchmarkEvalOpen-10       	      9	118168319 ns/op	2209928 B/op	  18928 allocs/op
BenchmarkEvalOpen-10       	      9	118805454 ns/op	2208294 B/op	  18924 allocs/op
BenchmarkEvalOpen-10       	      9	118506347 ns/op	2208712 B/op	  18922 allocs/op
BenchmarkEvalOpen-10       	      9	118898060 ns/op	2210256 B/op	  18926 allocs/op
BenchmarkEvalOpen-10       	      9	118450250 ns/op	2208210 B/op	  18924 allocs/op
BenchmarkEvalOpen-10       	      9	117723833 ns/op	2207528 B/op	  18922 allocs/op
BenchmarkEvalOpen-10       	      9	117134787 ns/op	2209227 B/op	  18925 allocs/op
BenchmarkEvalOpen-10       	      9	116210269 ns/op	2208843 B/op	  18926 allocs/op
BenchmarkEvalOpen-10       	      9	116987444 ns/op	2208736 B/op	  18925 allocs/op
BenchmarkEvalEnvLoad-10    	      4	298021334 ns/op	2216058 B/op	  18951 allocs/op
BenchmarkEvalEnvLoad-10    	      4	302557979 ns/op	2213974 B/op	  18944 allocs/op
BenchmarkEvalEnvLoad-10    	      4	293050229 ns/op	2212098 B/op	  18945 allocs/op
BenchmarkEvalEnvLoad-10    	      4	304410510 ns/op	2211322 B/op	  18946 allocs/op
BenchmarkEvalEnvLoad-10    	      4	301698562 ns/op	2212554 B/op	  18947 allocs/op
BenchmarkEvalEnvLoad-10    	      4	299588854 ns/op	2214102 B/op	  18946 allocs/op
BenchmarkEvalEnvLoad-10    	      4	295087740 ns/op	2211650 B/op	  18944 allocs/op
BenchmarkEvalEnvLoad-10    	      4	295875531 ns/op	2212638 B/op	  18950 allocs/op
BenchmarkEvalEnvLoad-10    	      4	294871781 ns/op	2212038 B/op	  18945 allocs/op
BenchmarkEvalEnvLoad-10    	      4	294592875 ns/op	2211682 B/op	  18945 allocs/op
BenchmarkEvalAll-10        	      3	405058722 ns/op	2215330 B/op	  18976 allocs/op
BenchmarkEvalAll-10        	      3	407002764 ns/op	2215688 B/op	  18978 allocs/op
BenchmarkEvalAll-10        	      3	409757153 ns/op	2214973 B/op	  18976 allocs/op
BenchmarkEvalAll-10        	      3	404553611 ns/op	2215261 B/op	  18977 allocs/op
BenchmarkEvalAll-10        	      3	402620945 ns/op	2216994 B/op	  18980 allocs/op
BenchmarkEvalAll-10        	      3	405302139 ns/op	2213112 B/op	  18973 allocs/op
BenchmarkEvalAll-10        	      3	404533556 ns/op	2215848 B/op	  18978 allocs/op
BenchmarkEvalAll-10        	      3	403431236 ns/op	2215896 B/op	  18979 allocs/op
BenchmarkEvalAll-10        	      3	402586597 ns/op	2217245 B/op	  18983 allocs/op
BenchmarkEvalAll-10        	      3	404775236 ns/op	2217122 B/op	  18980 allocs/op