Dispatch: replace runtime selector computation with compile-time

[mbenke]

Method selectors (4-byte function identifiers used for EVM dispatch) were computed at runtime: ContractDispatch.hs generated ABIString instances that emitted Yul assembly to build ABI signature strings in memory byte-by-byte, then hashed them with keccak256. This was complex and wasteful since signature strings are fully known at compile time.

This PR replaces the runtime approach with compile-time computation using the SigString typeclass. ContractDispatch.hs now generates SigString instances that return the method name as a string literal. dispatch.solc builds the full signature via string concatenation (concatLit) and hashes it with keccakLit - the partial evaluator folds the entire chain to an integer constant, so no string operations or hashing remain at runtime.

Changes

• ContractDispatch.hs: mkNameInst generates SigString instances (single sigStr method returning a StrLit) instead of ABIString instances (Yul byte-manipulation assembly). Removed generateStoreBytes, storeChunk, chunk32, bsToInteger and their imports.
• std/dispatch.solc: Replaced with ABIString the SigString-based dispatch
• std/rtdispatch.solc: Old runtime ABIString dispatch preserved for reference.
• std/std.solc: Added string:Add instance delegating to concatLit.

Testing

./run_contests.sh runs tests from test/examples/dispatch ensuring dispatch works correctly after changes (including miniERC20.solc, which is the main litmus test).

Potential TODO

Replace chained ifs by a big match - this would require allowing comptime expressions in patterns - TBD

[rodrigogribeiro]

Is there a reason for this file be in this commit or is it a temporary generated during tests?

[rodrigogribeiro]

Is this file necessary in this commit or is it a temporary?

[mbenke]

test/examples/dispatch/*.json files are used by run_contests.sh which is part of #314 - I forgot this has not been merged yet

[rodrigogribeiro]

No worries... after checking the other PR I noticed that. Sorry for the noise.

[rodrigogribeiro]

Is this option enabled by default? I'd argue that it would be useful to reduce the noise of the compiler output when using the standard library.

[mbenke]

No, it is enabled by the --dump-spec flag (so this is not printed even with -v)

[rodrigogribeiro]

Great work! LGTM.

[Y-Nak]

LGTM!