feat(compiler): update compiler profiles (#1562)

* Disable debug for benchmark profile

The benchmark profile previously had the debug option enabled. This commit comments out this section in the 'Cargo.toml' file, thus disabling the debug option. We want to avoid unnecessary debugging overhead during benchmarking tests:

1. code optimization: loop unrolling, inline, dead code elimination
2. extra instructions and data: less cache hit
3. branch prediction

only enable debug when error happened

* remove useless cli profile from Cargo.toml

1. The debug mode from the performance profile section has been removed in order to optimize runtime performance.

2. cli profile block has also been entirely removed from Cargo.toml, to simplify the configuration as it was redundant and unecessary.

* add comment about Enable overflow checks in release profile

The overflow checks have been activated in the release profile of the Cargo.toml file. This change doesn't significantly affect performance since branch prediction handles it quite efficiently. (actually almost no effect)

* Update Cargo.toml profiles configuration

Removed the non-recommended 'performance' profile from the Cargo.toml and added link-time optimization (LTO) to the 'release' and 'bench' profiles. The thin version of LTO has been used for its balance between performance improvement and compile time cost. Also, inherited the 'release' profile to the 'bench' profile for consistent configurations.

Author: popcnt1

Cargo.toml

@@ -336,26 +336,14 @@ move-ir-types = { git = "https://github.com/rooch-network/move", rev = "1b49c6e1
 # move-bytecode-source-map  = { path = "../move/language/move-ir-compiler/move-bytecode-source-map" }
 # move-ir-types = { path = "../move/language/move-ir/types" }
 [profile.release]
-debug = false
-overflow-checks = true
-
-# The performance build is not currently recommended
-# for production deployments. It has not been widely tested.
-[profile.performance]
-inherits = "release"
-opt-level = 3
-debug = true
+# enable overflow checks won't affect performance much, branch prediction will handle it well
 overflow-checks = true
+# enable link-time optimization, which can significantly improve perf but donesn't cost much extra in compile time
+# thin LTO is enough for us:
+# (https://blog.llvm.org/2016/06/thinlto-scalable-and-incremental-lto.html)
 lto = "thin"
-codegen-units = 1
-
-[profile.cli]
-inherits = "release"
-debug = false
-opt-level = "z"
-lto = "thin"
-strip = true
-codegen-units = 1
+codegen-units = 1 # Help to achieve better result with lto
 
 [profile.bench]
-debug = true
+inherits = "release"
+#debug = true