Text that grows during parsing can overflow Parse_text #7870

Problem

read_raw_file_text sizes Parse_text and Parse_text_raw from the file's length (allocate_parse_text(file_len + 1) in code/parse/parselo.cpp), but the text written into them can be longer than the file:

• In the default (non-Unicode) text mode, process_raw_file_text passes each line through maybe_convert_foreign_characters, which writes every ß (0xDF) as the two letters ss. That copy into Parse_text has no size check.
• A file using the older 7-bit encryption is decrypted into Parse_text_raw, and that format unpacks every 7 stored bytes into 8 characters.

So a table made mostly of ß characters, or an old-style encrypted table, can write past the end of the heap buffer when it is the largest file parsed so far.

Reproduction

Found by reading the code while reviewing a new text reader for #7865; I haven't triggered it in game, since the buffers only grow and a larger file read earlier hides it. Running the same line-splitting, comment-stripping and maybe_convert_foreign_characters code over a 1 MB file of 0xDF bytes (a newline every 100 bytes) wrote 2,086,667 bytes, about twice the allocation allocate_parse_text would make for that file.

Expected vs actual

Expected: the converted text fits its buffer.
Actual: the conversion can write past the end of Parse_text, and decryption past Parse_text_raw.

Fix alternatives

• Size both buffers for the worst case: twice the raw text in non-Unicode mode, 8/7 of the file for 7-bit encrypted files.
• Or bound the copy in process_raw_file_text and grow the buffer when it would overflow.

The .eff reader has the same problem with fixed stack buffers: #7869.