fix(highlight): env -S payload is a full arg list, not a bare interpreter

The attached split-string payload (`-Spython3`, `--split-string=...`) can
itself begin with env options or VAR=value assignments before the
interpreter: `-S-i python3`, `-SFOO=bar python3`,
`--split-string=-u FOO python3`. Rather than taking the payload's first
word as the interpreter, re-inject the attached payload into the token
stream so it flows through the same option / operand / assignment state
machine as a separated payload. Adds the three cases as resolver tests.

builtin/runtime/syntax.lua

@@ -98,13 +98,21 @@ function pmacs.parse.language_from_shebang(buf)
     base = nil
     local seen_env = false
     local skip_next = false
+    local tokens = {}
     for tok in rest:gmatch("%S+") do
-      -- `-S`/`--split-string` introduces a string whose FIRST word is the
+      tokens[#tokens + 1] = tok
-      -- interpreter. GNU env accepts that value ATTACHED — `-Spython3`,
+    end
-      -- `-vSpython3` (after no-operand short flags i/v/0), or
+    local i = 1
-      -- `--split-string=python3` — where the interpreter rides inside the
+    while i <= #tokens do
-      -- option token. Extract it; the separated forms (`-S python3`) are
+      local tok = tokens[i]
-      -- handled by simply walking on to the next token.
+      i = i + 1
+      -- `-S`/`--split-string` introduces a complete env argument list,
+      -- not necessarily an interpreter first: it may begin with more env
+      -- options or `VAR=value` assignments. GNU env accepts that value
+      -- ATTACHED — `-Spython3`, `-vSpython3` (after no-operand short flags
+      -- i/v/0), or `--split-string=python3`. Put an attached payload back
+      -- into this token stream so it goes through the same option/operand/
+      -- assignment state machine as a separated payload.
       local split_attached =
         tok:match("^%-[iv0]*S(.+)$") or tok:match("^%-%-split%-string=(.+)$")
       if not seen_env then
@@ -112,12 +120,10 @@ function pmacs.parse.language_from_shebang(buf)
       elseif skip_next then
         skip_next = false -- the operand consumed by the previous option
       elseif split_attached then
-        local word = split_attached:match("^(%S+)")
+        table.insert(tokens, i, split_attached)
-        base = word and (word:match("([^/]+)$") or word)
-        break
       elseif tok == "-S" or tok == "--split-string" then
         -- Separated split-string: the next token starts the string, i.e.
-        -- the interpreter — keep walking.
+        -- another env argument — keep walking.
       elseif tok:find("=", 1, true) then
         -- `VAR=value` env assignment, or another `--long=value` option:
         -- self-contained, skip.

tests/m4_acceptance.rs

@@ -5883,6 +5883,11 @@ fn m4_shebang_resolver_maps_interpreters() {
         ("#!/usr/bin/env -Spython3 -u\n", "python"),
         ("#!/usr/bin/env --split-string=python3 -u\n", "python"),
         ("#!/usr/bin/env -vSpython3 -u\n", "python"),
+        // The attached split string is a complete env argument list, so
+        // options and assignments may precede the interpreter within it.
+        ("#!/usr/bin/env -S-i python3 -u\n", "python"),
+        ("#!/usr/bin/env -SFOO=bar python3 -u\n", "python"),
+        ("#!/usr/bin/env --split-string=-u FOO python3\n", "python"),
         // GNU-env options that consume an operand must not have the
         // operand mistaken for the interpreter.
         ("#!/usr/bin/env -u FOO python3\n", "python"),