Normative: Arbitrary module namespace identifier names

[Jack-Works]

Implementing tc39/ecma262#2154

[Jack-Works]

Made a full rewrite of this PR. Just get the compiler part to work. Will handle language service days later.

[ExE-Boss]

Nit: extraneous whitespace:

Suggested change

	if (nameArg) {
	if (nameArg) {

[Jack-Works]

Hi, @DanielRosenwasser I have a question. With this PR, it's possible to define members with names that cannot be an identifier on a Symbol.

export const enum InternalSymbolName {
    Call = "__call", // Call signatures
    Constructor = "__constructor", // Constructor implementations
    New = "__new", // Constructor signatures
    Index = "__index", // Index signatures
    ExportStar = "__export", // Module export * declarations
    Global = "__global", // Global self-reference
    Missing = "__missing", // Indicates missing symbol
    Type = "__type", // Anonymous type literal symbol
    Object = "__object", // Anonymous object literal declaration
    JSXAttributes = "__jsxAttributes", // Anonymous JSX attributes object literal declaration
    Class = "__class", // Unnamed class expression
    Function = "__function", // Unnamed function expression
    Computed = "__computed", // Computed property name declaration with dynamic name
    Resolving = "__resolving__", // Indicator symbol used to mark partially resolved type aliases
    ExportEquals = "export=", // Export assignment symbol
    Default = "default", // Default export symbol (technically not wholly internal, but included here for usability)
    This = "this",
}

I wonder if this PR gonna make programmers able to write code that breaks TS by declaring a symbol member that has the same name as an internal symbol, like

export { item as "this" }
// or
export { item as "export=", item as __new }

[DanielRosenwasser]

I think the comment

Default export symbol (technically not wholly internal, but included here for usability)

is the key here, and the same applies to This. those symbols are just for comparisons in various places. If anything relied on this or default being special entities within a symbol table, they'd probably already be broken.

Maybe @weswigham or @gabritto can back me up on that statement though.