Handle DWARF string sections larger than 2 GB

[Martin Řezník (martin-reznik)]

Promotes the DWARF >2 GB string-section handling into main. Staged on fix/large-debug-str via #1207 (approved) and now being promoted.

What this does

• Adds an IDwarfStringReader abstraction for DWARF string sections, with a file-backed implementation (DwarfFileStringReader) that reads individual null-terminated strings directly from the binary instead of materializing the whole .debug_str section into a byte[].
• Adds the opt-in --dwarf-string-read-threshold CLI option. File-backed reads are used only when this option is supplied and the .debug_str section reaches the configured size.
• Preserves the existing in-memory behavior by default and for sections below the threshold.
• Adds regression tests for >2 GB reads and for the opt-in threshold wiring.

Root cause

ELFSharp materializes section contents into a byte[]; a .debug_str section larger than int.MaxValue overflows the 32-bit size conversion, preventing BinSkim from analyzing otherwise-valid ELF binaries (hit by Minecraft's Nintendo Switch symbol files).

Behavior / compatibility

• Fully opt-in. With no --dwarf-string-read-threshold, the value is null and every existing user keeps the exact in-memory code path (asserted by a unit test).
• No thread-safety change. Per-binary DWARF parsing is sequential and each ElfBinary builds/disposes its own reader per Lazy, so the stateful file-backed reader is never shared across BinSkim's file-level parallelism.
• Mach-O is adapted to the new interface (signature only); the threshold is wired through ElfBinary only.

Known limitations (non-blocking, tracked for later)

• Addressable offset is bounded at 4 GB (covers DWARF32 2-4 GB; not a >4 GB DWARF64 .debug_str).
• The opt-in read is byte-by-byte, so it is slower on very large sections.
• A >2 GB Mach-O .debug_str is still unsupported.

Refs

• Promotes Handle DWARF string sections larger than 2 GB #1207 by David Quiceno (@v-davidquiceno-ascendion) (Minecraft Engineering Systems / Ascendion).

[Alexander Köplinger (akoeplinger)]

looks like this is failing on some dotnet format errors?

[David Quiceno (v-davidquiceno-ascendion)]

Alexander Köplinger (@akoeplinger) Martin Řezník (@martin-reznik) I fixed the CRLF formatting issue in #1218. dotnet format src --verify-no-changes passes with the update.