Simplify documentation indexer

Write the index entirely in RAM. Its only a few hundred KiB, which
trivially fits in memory.

Compress it twice in memory, once to build the ZIP that is unpacked
at runtime, and again to package it into a JAR for linking with
the runtime. This saves a build step in the BUCK rules.

Move the ZIP under the server package name, to reduce any risk
of collision with another concept of "index.zip".

Change-Id: I74e59712e9855ac79c5220ff0a6b30ecbc3d152f

Documentation/BUCK

@@ -62,8 +62,7 @@ python_binary(
 genrule(
   name = 'index',
   cmd = '$(exe //lib/asciidoctor:doc_indexer) ' +
-      '-z $OUT ' +
-      '--tmp $TMP ' +
+      '-o $OUT ' +
       '--prefix "%s/" ' % DOC_DIR +
       '--in-ext ".txt" ' +
       '--out-ext ".html" ' +
@@ -73,20 +72,12 @@ genrule(
     ':licenses.txt',
     '//lib/asciidoctor:doc_indexer',
   ],
-  out = 'index.zip',
-)
-
-genrule(
-  name = 'index_jar',
-  cmd = 'jar cf $OUT -C $SRCDIR index.zip',
-  srcs = [genfile('index.zip')],
-  deps = [':index'],
   out = 'index.jar',
 )
 
 prebuilt_jar(
   name = 'index_lib',
   binary_jar = genfile('index.jar'),
-  deps = [':index_jar'],
+  deps = [':index'],
   visibility = ['PUBLIC'],
 )