gh-153569: initialize file encoding without tokenizing

Encoding lookup only needs the BOM and first two physical lines. Run the reader
initialization step directly instead of creating tokens and suppressing warnings
during unrelated work.

With lookup no longer entering the lexer, warning suppression is unnecessary
and ordinary tokenizer warnings keep a single path.

Author: pablogsal

Parser/lexer/state.h

@@ -90,14 +90,11 @@ struct tok_state {
     /* Stuff for PEP 0263 */
     char *encoding;         /* Source encoding. */
     const char* line_start;     /* pointer to start of current line */
-
     _PyTok_SourceText source;
     struct _PyTok_Reader *reader;
 
     int type_comments;      /* Whether to look for type comments */
 
-    int report_warnings;
-    // TODO: Factor this into its own thing
     tokenizer_mode tok_mode_stack[MAXFSTRINGLEVEL];
     int tok_mode_stack_index;
     int tok_extra_tokens;

Parser/tokenizer/helpers.c

@@ -98,10 +98,6 @@ _PyTokenizer_indenterror(struct tok_state *tok)
 int
 _PyTokenizer_warn_invalid_escape_sequence(struct tok_state *tok, int first_invalid_escape_char)
 {
-    if (!tok->report_warnings) {
-        return 0;
-    }
-
     PyObject *msg = PyUnicode_FromFormat(
         "\"\\%c\" is an invalid escape sequence. "
         "Such sequences will not work in the future. "
@@ -187,10 +183,6 @@ _PyTokenizer_raise_init_error(PyObject *filename)
 int
 _PyTokenizer_parser_warn(struct tok_state *tok, PyObject *category, const char *format, ...)
 {
-    if (!tok->report_warnings) {
-        return 0;
-    }
-
     PyObject *errmsg;
     va_list vargs;
     va_start(vargs, format);

Parser/tokenizer/reader.c

@@ -5,7 +5,6 @@
 #include "helpers.h"
 #include "reader.h"
 #include "reader_internal.h"
-#include "../lexer/lexer.h"
 #include "../lexer/state.h"
 
 #ifdef HAVE_UNISTD_H
@@ -865,13 +864,10 @@ _PyTokenizer_FindEncodingFilename(int fd, PyObject *filename)
         _PyTokenizer_Free(tok);
         return NULL;
     }
-    /* Reporting a warning here could recursively ask for the encoding. */
-    tok->report_warnings = 0;
-    while (tok->lineno < 2 && tok->done == E_OK) {
-        struct token token;
-        _PyToken_Init(&token);
-        _PyTokenizer_Get(tok, &token);
-        _PyToken_Free(&token);
+    if (initialize_file(tok) < 0) {
+        fclose(fp);
+        _PyTokenizer_Free(tok);
+        return NULL;
     }
     fclose(fp);
     char *encoding = tok->encoding == NULL