lang: lex: add support for kebab-case identifiers and negative numbers

identifiers can now contain hyphens, with the following restrictions:
 * an identifier cannot start or end with a hyphen.
 * an identifier cannot contain more than one hyphen in a row.

kebab-case identifiers can be used for type and variable names, as well
as message identifiers and labels.

to avoid ambiguity, the lexer now enforces whitespace around most binary
operators (with a few exceptions, such as semicolons). trying to
compile a "compact" arithmetic expression, such as

	y=1+2

will now result in a "missing whitespace" error.

lang/lex.h

@@ -8,12 +8,17 @@
 #include <ivy/status.h>
 #include <stdint.h>
 
+enum lex_token_flags {
+	LEX_TOK_REQUIRES_WHITESPACE = 0x01u,
+};
+
 struct ivy_lexer {
 	struct ivy_lexer_symbol_node *lex_sym_tree;
 	struct ivy_diag_ctx *lex_diag_ctx;
 	struct ivy_line_source *lex_source;
 	b_dict *lex_keywords;
 	enum ivy_status lex_status;
+	int lex_prev_char, lex_cur_char;
 
 	b_queue lex_queue;
 	enum ivy_token_type lex_prev_token;
@@ -47,7 +52,7 @@ struct lexer_state {
 
 struct ivy_lexer_symbol_node {
 	char s_char;
-	enum ivy_symbol s_id;
+	struct lex_token_def *s_def;
 
 	b_queue_entry s_entry;
 	b_queue s_children;
@@ -55,6 +60,7 @@ struct ivy_lexer_symbol_node {
 
 struct lex_token_def {
 	int id;
+	enum lex_token_flags flags;
 	const char *name;
 	uint64_t name_hash;
 };