vulkan: implement compute shaders

A lot of stuff probably isn't implemented yet, however the
following code already works:

```lua
local shader = love.graphics.newComputeShader [[
    layout (local_size_x = 1, local_size_y = 1) in;

    layout(r32f) uniform highp image2D out_tex;

    uniform float time;

    void computemain() {
        imageStore(out_tex, ivec2(love_GlobalThreadID.xy), vec4(time, 1.0, 1.0, 1.0));
    }
]]

local tex = love.graphics.newTexture(64, 64, {computewrite = true,format = "r32f"})

function love.draw()
    shader:send("out_tex", tex)
    shader:send("time", love.timer.getTime() % 1)
    love.graphics.dispatchThreadgroups(shader, 64, 64, 1)

    love.graphics.print("compute shader test")
    love.graphics.draw(tex, 25, 25)
end
```

src/modules/graphics/vulkan/Shader.h

@@ -25,11 +25,13 @@ public:
 	bool loadVolatile() override;
 	void unloadVolatile() override;
 
+	VkPipeline getComputePipeline() const;
+
 	const std::vector<VkPipelineShaderStageCreateInfo>& getShaderStages() const;
 
 	const VkPipelineLayout getGraphicsPipelineLayout() const;
 
-	void cmdPushDescriptorSets(VkCommandBuffer, uint32_t currentFrame);
+	void cmdPushDescriptorSets(VkCommandBuffer, uint32_t currentFrame, VkPipelineBindPoint);
 
 	void attach() override;
 
@@ -74,6 +76,8 @@ private:
 
 	VkDeviceSize uniformBufferSizeAligned;
 
+	VkPipeline computePipeline;
+
 	VkDescriptorSetLayout descriptorSetLayout;
 	VkPipelineLayout pipelineLayout;
 
@@ -90,6 +94,8 @@ private:
 	Graphics* gfx;
 	VkDevice device;
 
+	bool isCompute = false;
+
 	std::unordered_map<std::string, graphics::Shader::UniformInfo> uniformInfos;
 	UniformInfo* builtinUniformInfo[BUILTIN_MAX_ENUM];