You cannot because those BGP routes were learned as l3-vpn unicast routes and inet.0 accepts BGP famiy inet routes.
What you can do is configure a static route in inet.0 for the BGP prefixes in the L3.inet.0, with next-table L3.inet.0 instead of a next-hop.
For example, if you are receiving prefixes from the remote PE for the L3 VPN, for subnets of 172.16/16, you can create a static route that summarizes those routes, and points to table L3.inet.0, and then redistribute it into OSPF as shown:
set routing-options static route 172.16/16 next-table L3.inet.0
set policy-options policy-statement static2ospf term 1 from protocol static
set policy-options policy-statement static2ospf term 1 from route-filter 172.16.0.0/16 exact
set policy-options policy-statement static2ospf term 1 from condition remote-prefixes ### so route is only advertised
set policy-options policy-statement static2ospf term 1 then accept ### if the present in L3.inet.0
set policy-options condition remote-prefixes if-route-exists 172.16.2.1/32
set policy-options condition remote-prefixes if-route-exists table L3.inet.0
set protocols ospf export static2ospf