Refactor codegen: split FunC AST generation and pretty-printing

[anton-trunov]

Some remarks on the design of the new codegen architecture:

• Code generation should be a two-step transformation: (1) produce FunC ASTs, (2) pretty-print FunC ASTs which we then use to feed into the FunC compiler as we do new;
• Special care should be taken to avoid clashes between identifiers and FunC keywords;
• The FunC pretty-printer should produce human readable code that should look like the currently produced FunC code;
• The FunC pretty-printer should have configuration options to support fully-parenthesized expressions and more human-readable version (we should focus on the easier fully-parenthesized version first);
• The TypeScript representation of FunC AST and the pretty-printer should be available as a separate NPM package so it can be re-used by TON community.

Related issue in the ongoing series of refactorings: #458.

[anton-trunov]

@byakuren-hijiri Btw, the pretty-printer and the FunC AST data structure should be adaptable to #420

[novusnota]

If we were to parse FunC via Ohm, FuncParser.bnf and FuncLexer.flex from ton-blockchain/intellij-ton can be helpful guidances as there's no formal description of FunC. Also, it's parser is almost exclusively located in one file, which is helpful too: parse-func.cpp

[anton-trunov]

Retracting the requirement

• The pretty-printer and the FunC AST data types itself should be adaptable to FunC v0.5 migration #420;

as not planned