[fidl] Compiler implementation for tables

Test: updated existing tests, implemented a table test suite for table edge cases

Change-Id: Id145d12f44544027d0d28c2288ebf5c6de39b446

[fidl][ftp-009] Adds Documentation comment desugaring to attribute

extends the JSON IR to support attributes on members of structured data types

Test: Migrate comments to doc comments and check the output of the json IR

Change-Id: Ib2d5e59682b49aa5103440fdec8228da00170848

[fidl] Remove |status| keyword

All clients have migrated to using zx.status, which is not a language
keyword.

Test: Compiles
Change-Id: I9c2c08b5c4062c598234f4f461b53a44e8aaf5a7

[fidl] Store attributes for methods.

The current syntax that is parsed allows methods to have attributes,
but it is not currently used anywhere.

This change includes those attributes in the JSON output for processing
by code generators.

TEST=fx build

Change-Id: Ibaef4871cc6110f9b55882869035b56d10ef5160

[fidl] Keep track of the maximum handles required to represent types.

This is motivated by the desire to restrict the Clone method to types
that cannot contain handles.

This adds maximum handle tracking in TypeShape. It's mostly fairly
straightforward but there are a couple of nastier edge cases.

First unbounded vectors could have any number of members. The number of
members are in practice bounded by the maximum message size, so the
number of handles for an unbounded vector is the number of handles
multiplied by the maximum message size.

Similarly recursive structs and unions are bounded by the message size.
The compiler now has to keep track of reentrant compilation.

Messages with handles in either unbounded vectors or recursive types
will be large.

Test: wrote some simple FIDL and verified that the IR JSON was correct.

Change-Id: Ibbe2f04a63774c6cafc9205180795ccbd2d2048c

[fidl] Allow libraries to be namespaced

Change-Id: I64143e7cc9aef8050f32409abc8574920ba444a1

[fidl] Fix spelling of attributes-list in the json ir

FIDL-155 #done

Change-Id: I599d9354fe2eebc3d7baff099f60da4fb9e2723e

[fidl][json] Introduce a schema version

Change-Id: I677fbf85a6c535aca0cc2b0fb607528899a30c02

[fidl] Generate library_dependencies in IR

Change-Id: Ic73c6da815b1a9d63bdfbf02d9b12e7fd1146486

[fidl] Name compound identifiers as a canonically encoded string

This makes it considerably easier to look them up in the go backend.

Change-Id: I9fc79f7ed869554e42e56d84da8a3ff2bd24ea8f

[fidl] A more correct compound identifier

Change-Id: I3b8fbbcc712094eb24a70332cbf2833d67bbe4a4

[fidl] Add support for attributes on declarations

We'll use this for the [ServiceName] attribute on interfaces.

Change-Id: I1371e6363fd1dc3566f18c8721de0ed3e6203634

[fidl] Fix Name json generation

Previously, this type was a raw::CompoundIdentifier. Eventually Names
will be fancy, and not always constructed with one part. Until then,
generate the list of always one part.

This also lines up the json schema with the flat::ast, in terms of
short vs long names.

Change-Id: Ia018b84b8d0c0aee465083c003cd390a69b19907

[fidl] Front-end should resolve element counts

Rather than making the backends translate the element counts from
identifiers to numerical values, the front-end now resolves the
identifiers and provides the numerical values.

TO-746 #done

Change-Id: I0ca6df2b06e43e2f260a7563aa2c50d6e991e5e9

[fidl][zx] Sync object types between FIDL and zx

Several of the object types referenced by the FIDL spec no longer exist.
Also, add two missing object types to libzx.

Change-Id: I18a00110912b99fd734a59e5a390de94ef433799

[fidl] Serialize all type sizes and alignments

TO-825

Change-Id: I09aabc7c52ebf28180470719d77b3f238e49c4d3

[fidl] Disallow empty structs

These are problematic to represent in C and C++. Most people who want
them can work around this with a dummy field, for now.

Change-Id: Ic5c6dba0d2a45a2e26193c3c4f4cb188f5e24c86

[fidl] Provide declaration order in the json representation

This information needs to be computed for C and C++ bindings, since it
needs to lay out structures in order of their inclusion.

This information gets computed at the same time that structure sizes
and offsets are computed, since that computation happens in the same
dependency order.

Change-Id: I9b28fbd9c5b6e50c3f43af50678390a14c01e1d4

[fidl] Add a type map to FIDL JSON IR

This map will make it easier for the backend to determine to what kind
of thing a type identifier refers.

TO-745 #done

Change-Id: Iae7f0f3ed4cae0caa676fc360040d77cdaf18979

[fidl] Add method sizes and offsets to the json

TO-758

Change-Id: Ie4a1d7cae536a35a7e5cd749bc0fe31f8a13ffc1

[fidl] Fix misplaced attribute in schema.

Change-Id: Idd738236fc96f5f726f898ffd4addf10f8b02318

[fidl] Add struct and union sizes and member offsets to the json

TO-758

Change-Id: I1fffbf4dd40edcbc888177ec85234f5cdd659afb

[fidl] Disallow the 0 ordinal in the json schema

Change-Id: Iac6ecbad32b21079d5c95998cde0d594a12358dc

[fidl][json] Make the schema valid json

Change-Id: Icd9fe288ae3acc5def7c6e871815dfe4bd1cb6f4

[fidl] Include the library's name in json generation

Change-Id: Ibd7c541f43f20a7b95e7edeb4b48fd8d35ce09a9

[fidl] Rename nullability to nullable in the json

Change-Id: I4f4c18d679dd94f74bafbb040fb9f0b6516fcd96

[fidl] Refer to libraries as such

After some churn, we settled on 'library' for a related, self
contained group of fidl files.

Change-Id: I9bcc4b0301f16247f67716d3ed1a1b67e5a387d2

[fidl] Simplify the enum type representation

Enum types are always integer types. We can just encode that directly
into JSON without needing all the structure that's required to encode an
arbitrary type.

Change-Id: I31053c75c7afff2f7d60adb4065f26d74f4109a3

[fidl] Generate JSON intermeidate representation

This JSON IR will be used by the various language backends to create
language-specific bindings.

Change-Id: I9555f43b7f7363a1adfbaf34d0a67eceeeb2a37d